Skip to content

armgit / Filee

Public Imported from GitHub
0

Free, open-source file converter. Convert 263 formats offline by drag & drop — images, PDF, Word, Excel, PowerPoint, HWP/HWPX, video, audio, e-books and archives. No upload, no Office needed.

feat(ocr): add Apple Vision native and Tesseract OCR engine support

smjrifle216282a748cd
diff --git a/docs/ENGINES.md b/docs/ENGINES.md
index a7c4ce25..38907d75 100644
--- a/docs/ENGINES.md
+++ b/docs/ENGINES.md
@@ -34,6 +34,8 @@ Filee chooses engines automatically. Settings → *Engines* shows their status,
| Ghostscript | **optional download** (~20 MB, 31 MB on disk) | EPS/PS → PDF, PDF → EPS/PS, PostScript-based AI → PDF (images through PDF and PDFium); smaller PDFs when a preset asks for it |
| FFmpeg | **optional download** (~100 MB, 272 MB on disk) | **all video and audio**: video ↔ video (H.264, HEVC, AV1), video → animated GIF / APNG, a still frame or every frame, audio and subtitle extraction, audio ↔ audio (also APE, WavPack, TTA, DSD, DTS, ALAC, …), GIF, APNG and photos → MP4/WEBM/MOV, Live / Motion Photos → their video, a clip of a video → an iPhone Live Photo or Android Motion Photo (experimental), images → HEIC |
| Calibre | **optional download** (~216 MB, 660 MB on disk) | rare e-book formats: LIT, LRF, CHM, PDB, PML, RB, SNB, TCR, OEB → EPUB; EPUB → MOBI, AZW3, LIT, LRF, PDB, PML, RB, SNB, TCR; e-books → PDF (ranked above the HWPX writer, see below) |
+| Apple Vision | part of macOS (macOS only) | **scanned PDFs and images → editable text / DOCX / Markdown / HWPX** via native neural OCR (Latin, CJK, Cyrillic); zero download size |
+| Tesseract | **optional, install yourself** (~30 MB + the trained data of each language) | OCR for scripts Apple Vision doesn't know: Devanagari (Nepali, Hindi, Sanskrit, Marathi), Cyrillic beyond Russian/Ukrainian, Thai, Greek, Arabic script and more; the only OCR on Windows and Linux (Linux has no built-in option, so this is required there) |
DOCX, XLSX, XLS, ODS and PPTX → PDF need neither Microsoft Office nor LibreOffice: they are read in-process, written as HWPX
and rendered by rhwp. Anything the built-in readers understand (including PDF) is written as DOCX by the DOCX writer.
diff --git a/src/Filee.App/Assets/i18n/ar.json b/src/Filee.App/Assets/i18n/ar.json
index 7ee866f4..c27034f0 100644
--- a/src/Filee.App/Assets/i18n/ar.json
+++ b/src/Filee.App/Assets/i18n/ar.json
@@ -140,6 +140,7 @@
"engine.ghostscript.description": "تحويل EPS وPostScript وملفات Illustrator القديمة إلى PDF وصور؛ وتحويل PDF إلى EPS وPS.",
"engine.calibre.description": "صيغ الكتب الإلكترونية النادرة (LIT وLRF وCHM وPDB وRB…) وإخراج Kindle ‏(MOBI وAZW3).\nعند وضعه فوق محرك كتابة HWPX، يُنشئ أيضًا PDF من الكتب الإلكترونية (بسرعة أكبر بكثير للكتب الطويلة).",
"engine.libreoffice.description": "قراءة الصيغ القديمة والأقل شيوعًا: DOC وPPT وPages وKeynote وNumbers وWordPerfect وClarisWorks وPublisher وVisio…\nوالحفظ بصيغ DOC وXLS وPPT وODT وODP وقوالب OpenDocument وEMF.",
+ "engine.image-ocr.description": "استخراج النص من ملفات PDF الممسوحة ضوئيًا والصور إلى مستندات قابلة للتحرير (Word والنص وHWPX).",
"profile.images": "الصور",
"profile.pdf": "PDF",
"profile.office": "المستندات",
@@ -519,6 +520,14 @@
"presets.leave_out_pictures": "استبعاد الصور (النص فقط)",
"presets.leave_out_pictures_hint": "عند التعطيل: تُحفَظ الصور في مجلد name_files بجانب ملف Markdown، وتُدرَج روابطها فيه.\nعند التفعيل: النص فقط، دون مجلد.",
"presets.target_folder": "مجلد (استخراج الأرشيفات)",
+ "presets.group.ocr": "التعرف البصري على الأحرف (OCR)",
+ "presets.ocr_enabled": "التعرف على النص في المستندات الممسوحة ضوئيًا والصور",
+ "presets.ocr_hint": "استخراج نص قابل للتحرير من الصفحات الممسوحة ضوئيًا والصور إلى Word والنص والمستندات.",
+ "presets.ocr_language": "لغة التعرف",
+ "presets.ocr_language_hint": "اللغة المستخدمة للتعرف بدقة على الأحرف والكلمات.",
+ "presets.ocr_language.auto": "تلقائي / الافتراضي للنظام",
+ "presets.ocr_engine_required": "يتطلب Tesseract. قم بتثبيته من الإعدادات ← المحركات",
+ "presets.ocr_disabled_linux": "لا يحتوي Linux على OCR مدمج. قم بتثبيت Tesseract من الإعدادات ← المحركات للتمكين.",
"presets.group.media": "الفيديو والصوت",
"presets.media_quality": "الجودة",
"presets.media_quality_hint": "عندما تستطيع الصيغة الجديدة الاحتفاظ بالترميزات كما هي (MOV من iPhone إلى MP4)، تتغير الحاوية فقط.\nدون فقدان جودة، مع الاحتفاظ بـ HDR، خلال ثوانٍ.\nخيار «صغير» يعيد الترميز دائمًا لتقليل الحجم.",
@@ -652,6 +661,11 @@
"engines.guidance.linux.libreoffice": "ثبّته باستخدام مدير الحزم (على Ubuntu/Debian: sudo apt install libreoffice)، ثم اختر «التحقق مجددًا».\nيتطلب فتح HWP أيضًا Java وإضافة H2Orestart.",
"engines.guidance.linux.ghostscript": "ثبّته باستخدام مدير الحزم (على Ubuntu/Debian: sudo apt install ghostscript)، ثم اختر «التحقق مجددًا».",
"engines.guidance.windows": "لا يوفر Filee تنزيلًا لـ {0} لهذا الحاسوب.\nاختر «استخدام نسختي…» وحدّد نسخة مثبّتة.",
+ "engines.package.tesseract.name": "Tesseract (التعرف على النصوص متعدد اللغات)",
+ "engines.package.tesseract.description": "يتعرف على النص بأكثر من 100 لغة، بما في ذلك الديفاناغارية (النيبالية/الهندية) والعربية والخطوط النادرة.\nعلى macOS، يعمل Apple Vision أصلاً للغات الإنجليزية ولغات CJK واللغات الأوروبية بدونه.",
+ "engines.guidance.macos.tesseract": "قم بتثبيت Tesseract باستخدام Homebrew (brew install tesseract tesseract-lang)، ثم اختر “تحقق مرة أخرى”.\nأو اختر “استخدام نسختي…” وحدد مجلد bin الخاص به.",
+ "engines.guidance.linux.tesseract": "قم بتثبيت Tesseract باستخدام مدير الحزم (Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all)، ثم اختر “تحقق مرة أخرى”.\nأو اختر “استخدام نسختي…” وحدد مجلد البرنامج الخاص به.",
+ "engines.guidance.windows.tesseract": "قم بتثبيت Tesseract من UB-Mannheim (github.com/UB-Mannheim/tesseract)، ثم اختر “تحقق مرة أخرى” أو حدد مجلد التثبيت الخاص به.",
"engines.use_own_app_folder": "اختر المجلد الذي يحتوي على {0}.app",
"engines.progress_detail": "{0} من {1} · {2}/ث · {3}",
"engines.eta.estimating": "جارٍ تقدير الوقت المتبقي…",
diff --git a/src/Filee.App/Assets/i18n/en.json b/src/Filee.App/Assets/i18n/en.json
index d6024c30..31353d40 100644
--- a/src/Filee.App/Assets/i18n/en.json
+++ b/src/Filee.App/Assets/i18n/en.json
@@ -147,6 +147,7 @@
"engine.ghostscript.description": "EPS, PostScript and older Illustrator files to PDF and images; PDF to EPS and PS.",
"engine.calibre.description": "Rare e-book formats (LIT, LRF, CHM, PDB, RB…) and Kindle output (MOBI, AZW3).\nAbove the HWPX writer, it also makes the PDFs of e-books (much faster for long books).",
"engine.libreoffice.description": "Reads older and less common formats: DOC, PPT, Pages, Keynote, Numbers, WordPerfect, ClarisWorks, Publisher, Visio…\nAlso saves as DOC, XLS, PPT, ODT, ODP, OpenDocument templates and EMF.",
+ "engine.image-ocr.description": "Extracts text from scanned PDFs and pictures into editable documents (Word, text, HWPX).",
"profile.images": "Images",
"profile.pdf": "PDF",
@@ -534,6 +535,14 @@
"presets.leave_out_pictures": "Leave out pictures (text only)",
"presets.leave_out_pictures_hint": "Off: the pictures are saved in a \"name_files\" folder next to the Markdown file and linked from it.\nOn: only the text, no folder.",
"presets.target_folder": "Folder (extract archives)",
+ "presets.group.ocr": "Optical Character Recognition (OCR)",
+ "presets.ocr_enabled": "Recognize text in scans and images",
+ "presets.ocr_hint": "Extracts editable text from scanned pages and images into Word, text, and documents.",
+ "presets.ocr_language": "Recognition language",
+ "presets.ocr_language_hint": "Language used to recognize characters and words accurately.",
+ "presets.ocr_language.auto": "Automatic / System Default",
+ "presets.ocr_engine_required": "Needs Tesseract. Install it in Settings → Engines",
+ "presets.ocr_disabled_linux": "Linux has no built-in OCR. Install Tesseract in Settings → Engines to enable.",
"presets.group.media": "Video & audio",
"presets.media_quality": "Quality",
"presets.media_quality_hint": "When the new format can hold the codecs as they are (iPhone MOV → MP4), the video is only rewrapped.\nNo quality lost, HDR kept, done in seconds.\nSmall always re-encodes to make the file smaller.",
@@ -669,6 +678,11 @@
"engines.guidance.linux.libreoffice": "Install it with your package manager (Ubuntu/Debian: sudo apt install libreoffice), then choose “Check again”.\nOpening HWP also needs Java and the H2Orestart extension.",
"engines.guidance.linux.ghostscript": "Install it with your package manager (Ubuntu/Debian: sudo apt install ghostscript), then choose “Check again”.",
"engines.guidance.windows": "Filee has no download of {0} for this PC.\nChoose “Use my copy…” and select an installed copy.",
+ "engines.package.tesseract.name": "Tesseract (multilingual text recognition)",
+ "engines.package.tesseract.description": "Recognizes text in over 100 languages, including Devanagari (Nepali/Hindi), Arabic, and rare scripts.\nOn macOS, Apple Vision works natively for English, CJK, and European languages without it.",
+ "engines.guidance.macos.tesseract": "Install Tesseract with Homebrew (brew install tesseract tesseract-lang), then choose “Check again”.\nOr choose “Use my copy…” and select its bin folder.",
+ "engines.guidance.linux.tesseract": "Install Tesseract with your package manager (Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all), then choose “Check again”.\nOr choose “Use my copy…” and select its program folder.",
+ "engines.guidance.windows.tesseract": "Install Tesseract from UB-Mannheim (github.com/UB-Mannheim/tesseract), then choose “Check again” or select its install folder.",
"engines.use_own_app_folder": "Choose the folder containing {0}.app",
"engines.progress_detail": "{0} of {1} · {2}/s · {3}",
"engines.eta.estimating": "estimating time left…",
diff --git a/src/Filee.App/Assets/i18n/es.json b/src/Filee.App/Assets/i18n/es.json
index 89007a3e..b3da5d10 100644
--- a/src/Filee.App/Assets/i18n/es.json
+++ b/src/Filee.App/Assets/i18n/es.json
@@ -140,6 +140,7 @@
"engine.ghostscript.description": "EPS, PostScript y archivos antiguos de Illustrator a PDF e imágenes; PDF a EPS y PS.",
"engine.calibre.description": "Formatos raros de libros electrónicos (LIT, LRF, CHM, PDB, RB…) y salida Kindle (MOBI, AZW3).\nPor encima del escritor HWPX, también crea los PDF de libros electrónicos (mucho más rápido para libros largos).",
"engine.libreoffice.description": "Lee formatos antiguos y poco habituales: DOC, PPT, Pages, Keynote, Numbers, WordPerfect, ClarisWorks, Publisher, Visio…\nTambién guarda como DOC, XLS, PPT, ODT, ODP, plantillas OpenDocument y EMF.",
+ "engine.image-ocr.description": "Extrae texto de PDF escaneados e imágenes a documentos editables (Word, texto, HWPX).",
"profile.images": "Imágenes",
"profile.pdf": "PDF",
"profile.office": "Documentos",
@@ -519,6 +520,14 @@
"presets.leave_out_pictures": "Omitir imágenes (solo texto)",
"presets.leave_out_pictures_hint": "Desactivado: las imágenes se guardan en una carpeta \"nombre_files\" junto al archivo Markdown y se enlazan desde él.\nActivado: solo el texto, sin carpeta.",
"presets.target_folder": "Carpeta (extraer archivos comprimidos)",
+ "presets.group.ocr": "Reconocimiento óptico de caracteres (OCR)",
+ "presets.ocr_enabled": "Reconocer texto en escaneos e imágenes",
+ "presets.ocr_hint": "Extrae texto editable de páginas escaneadas e imágenes a Word, texto y documentos.",
+ "presets.ocr_language": "Idioma de reconocimiento",
+ "presets.ocr_language_hint": "Idioma utilizado para reconocer caracteres y palabras con precisión.",
+ "presets.ocr_language.auto": "Automático / Predeterminado del sistema",
+ "presets.ocr_engine_required": "Requiere Tesseract. Instálalo en Configuración → Motores",
+ "presets.ocr_disabled_linux": "Linux no tiene OCR integrado. Instala Tesseract en Configuración → Motores para activarlo.",
"presets.group.media": "Vídeo y audio",
"presets.media_quality": "Calidad",
"presets.media_quality_hint": "Cuando el formato nuevo puede conservar los códecs tal como están (MOV de iPhone → MP4), el vídeo solo se cambia de contenedor sin recodificar.\nSin pérdida de calidad, HDR conservado, listo en segundos.\nPequeño siempre recodifica para que el archivo pese menos.",
@@ -652,6 +661,11 @@
"engines.guidance.linux.libreoffice": "Instálalo con tu gestor de paquetes (Ubuntu/Debian: sudo apt install libreoffice) y luego elige «Comprobar otra vez».\nAbrir HWP necesita además Java y la extensión H2Orestart.",
"engines.guidance.linux.ghostscript": "Instálalo con tu gestor de paquetes (Ubuntu/Debian: sudo apt install ghostscript) y luego elige «Comprobar otra vez».",
"engines.guidance.windows": "Filee no tiene una descarga de {0} para este PC.\nElige «Usar mi copia…» y selecciona una copia instalada.",
+ "engines.package.tesseract.name": "Tesseract (reconocimiento de texto multilingüe)",
+ "engines.package.tesseract.description": "Reconoce texto en más de 100 idiomas, incluidos devanagari (nepalí/hindi), árabe y alfabetos poco comunes.\nEn macOS, Apple Vision funciona de forma nativa para inglés, CJK e idiomas europeos sin él.",
+ "engines.guidance.macos.tesseract": "Instala Tesseract con Homebrew (brew install tesseract tesseract-lang), luego elige “Comprobar de nuevo”.\nO elige “Usar mi copia…” y selecciona su carpeta bin.",
+ "engines.guidance.linux.tesseract": "Instala Tesseract con el gestor de paquetes (Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all), luego elige “Comprobar de nuevo”.\nO elige “Usar mi copia…” y selecciona su carpeta de programas.",
+ "engines.guidance.windows.tesseract": "Instala Tesseract desde UB-Mannheim (github.com/UB-Mannheim/tesseract), luego elige “Comprobar de nuevo” o selecciona su carpeta de instalación.",
"engines.use_own_app_folder": "Elige la carpeta que contiene {0}.app",
"engines.progress_detail": "{0} de {1} · {2}/s · {3}",
"engines.eta.estimating": "calculando el tiempo restante…",
diff --git a/src/Filee.App/Assets/i18n/hi.json b/src/Filee.App/Assets/i18n/hi.json
index dde78607..deaca383 100644
--- a/src/Filee.App/Assets/i18n/hi.json
+++ b/src/Filee.App/Assets/i18n/hi.json
@@ -140,6 +140,7 @@
"engine.ghostscript.description": "EPS, PostScript और पुरानी Illustrator फ़ाइलें PDF और इमेज में; PDF से EPS और PS।",
"engine.calibre.description": "दुर्लभ ई-बुक फ़ॉर्मैट (LIT, LRF, CHM, PDB, RB…) और Kindle आउटपुट (MOBI, AZW3)।\nHWPX writer के ऊपर होने पर यह ई-बुक के PDF भी बनाता है (लंबी किताबों के लिए बहुत तेज़)।",
"engine.libreoffice.description": "पुराने और कम प्रचलित फ़ॉर्मैट पढ़ता है: DOC, PPT, Pages, Keynote, Numbers, WordPerfect, ClarisWorks, Publisher, Visio…\nDOC, XLS, PPT, ODT, ODP, OpenDocument टेम्पलेट और EMF के रूप में सहेजता भी है।",
+ "engine.image-ocr.description": "स्कैन किए गए PDF और चित्रों से टेक्स्ट निकाल कर संपादन योग्य दस्तावेज़ों (Word, टेक्स्ट, HWPX) में बदलता है।",
"profile.images": "इमेज",
"profile.pdf": "PDF",
"profile.office": "दस्तावेज़",
@@ -519,6 +520,14 @@
"presets.leave_out_pictures": "तस्वीरें छोड़ दें (केवल टेक्स्ट)",
"presets.leave_out_pictures_hint": "बंद: तस्वीरें Markdown फ़ाइल के बगल में \"name_files\" फ़ोल्डर में सहेजी जाती हैं और उसी से लिंक होती हैं।\nचालू: केवल टेक्स्ट, कोई फ़ोल्डर नहीं।",
"presets.target_folder": "फ़ोल्डर (आर्काइव निकालने के लिए)",
+ "presets.group.ocr": "ऑप्टिकल कैरेक्टर रिकग्निशन (OCR)",
+ "presets.ocr_enabled": "स्कैन और छवियों में टेक्स्ट पहचानें",
+ "presets.ocr_hint": "स्कैन किए गए पेजों और छवियों से संपादन योग्य टेक्स्ट निकाल कर Word, टेक्स्ट और दस्तावेज़ों में बदलें।",
+ "presets.ocr_language": "पहचान भाषा",
+ "presets.ocr_language_hint": "वर्णों और शब्दों को सटीक रूप से पहचानने के लिए उपयोग की जाने वाली भाषा।",
+ "presets.ocr_language.auto": "स्वचालित / सिस्टम डिफ़ॉल्ट",
+ "presets.ocr_engine_required": "Tesseract की आवश्यकता है। इसे सेटिंग्स → इंजन में इंस्टॉल करें",
+ "presets.ocr_disabled_linux": "Linux में कोई अंतर्निहित OCR नहीं है। सक्षम करने के लिए सेटिंग्स → इंजन में Tesseract इंस्टॉल करें।",
"presets.group.media": "वीडियो और ऑडियो",
"presets.media_quality": "क्वालिटी",
"presets.media_quality_hint": "जब नया फ़ॉर्मैट कोडेक को वैसे ही रख सकता है (iPhone MOV → MP4), तो वीडियो केवल दोबारा लपेटा जाता है।\nकोई क्वालिटी नहीं घटती, HDR बना रहता है, और कुछ सेकंड में हो जाता है।\nछोटा विकल्प फ़ाइल घटाने के लिए हमेशा दोबारा एन्कोड करता है।",
@@ -652,6 +661,11 @@
"engines.guidance.linux.libreoffice": "इसे अपने पैकेज मैनेजर से इंस्टॉल करें (Ubuntu/Debian: sudo apt install libreoffice), फिर “फिर से जाँचें” चुनें।\nHWP खोलने के लिए Java और H2Orestart एक्सटेंशन भी चाहिए।",
"engines.guidance.linux.ghostscript": "इसे अपने पैकेज मैनेजर से इंस्टॉल करें (Ubuntu/Debian: sudo apt install ghostscript), फिर “फिर से जाँचें” चुनें।",
"engines.guidance.windows": "इस PC के लिए Filee के पास {0} का डाउनलोड नहीं है।\n“मेरी कॉपी इस्तेमाल करें…” चुनें और इंस्टॉल की गई कॉपी चुनें।",
+ "engines.package.tesseract.name": "Tesseract (बहुभाषी टेक्स्ट पहचान)",
+ "engines.package.tesseract.description": "देवनागरी (नेपाली/हिंदी), अरबी और दुर्लभ लिपियों सहित 100 से अधिक भाषाओं में टेक्स्ट पहचानता है।\nmacOS पर Apple Vision अंग्रेजी, CJK और यूरोपीय भाषाओं के लिए बिना इसके भी काम करता है।",
+ "engines.guidance.macos.tesseract": "Homebrew से Tesseract इंस्टॉल करें (brew install tesseract tesseract-lang), फिर “दोबारा जांचें” चुनें।\nया “मेरी कॉपी इस्तेमाल करें…” चुनें और इसके bin फ़ोल्डर का चयन करें।",
+ "engines.guidance.linux.tesseract": "अपने पैकेज मैनेजर से Tesseract इंस्टॉल करें (Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all), फिर “दोबारा जांचें” चुनें।\nया “मेरी कॉपी इस्तेमाल करें…” चुनें और इसका प्रोग्राम फ़ोल्डर चुनें।",
+ "engines.guidance.windows.tesseract": "UB-Mannheim (github.com/UB-Mannheim/tesseract) से Tesseract इंस्टॉल करें, फिर “दोबारा जांचें” चुनें या इसके इंस्टाल फ़ोल्डर का चयन करें।",
"engines.use_own_app_folder": "वह फ़ोल्डर चुनें जिसमें {0}.app है",
"engines.progress_detail": "{1} में से {0} · {2}/s · {3}",
"engines.eta.estimating": "बचा हुआ समय आँका जा रहा है…",
diff --git a/src/Filee.App/Assets/i18n/ja.json b/src/Filee.App/Assets/i18n/ja.json
index afc5f5d3..bc537324 100644
--- a/src/Filee.App/Assets/i18n/ja.json
+++ b/src/Filee.App/Assets/i18n/ja.json
@@ -140,6 +140,7 @@
"engine.ghostscript.description": "EPS、PostScript、古いIllustratorファイルをPDFや画像に変換し、PDFをEPSやPSに変換します。",
"engine.calibre.description": "一般的でない電子書籍形式(LIT、LRF、CHM、PDB、RBなど)を変換し、Kindle用のMOBI、AZW3形式で保存できます。\n「HWPX writer」より上に置くと、電子書籍のPDF作成もCalibreが担当します(長い本では大幅に速くなります)。",
"engine.libreoffice.description": "古い形式や一般的でない形式(DOC、PPT、Pages、Keynote、Numbers、WordPerfect、ClarisWorks、Publisher、Visioなど)を読み込みます。\nDOC、XLS、PPT、ODT、ODP、OpenDocumentテンプレート、EMF形式での保存にも対応します。",
+ "engine.image-ocr.description": "スキャンしたPDFや画像からテキストを抽出し、編集可能な文書(Word、テキスト、HWPX)に変換します。",
"profile.images": "画像",
"profile.pdf": "PDF",
"profile.office": "文書",
@@ -519,6 +520,14 @@
"presets.leave_out_pictures": "画像を含めない(テキストのみ)",
"presets.leave_out_pictures_hint": "オフ:画像をMarkdownファイルと同じ場所の「name_files」フォルダーに保存し、Markdownファイルからリンクします。\nオン:テキストだけを保存し、フォルダーは作成しません。",
"presets.target_folder": "フォルダー(アーカイブを展開)",
+ "presets.group.ocr": "光学式文字認識 (OCR)",
+ "presets.ocr_enabled": "スキャンや画像内のテキストを認識",
+ "presets.ocr_hint": "スキャンしたページや画像から編集可能なテキストを抽出し、Word、テキスト、ドキュメントに変換します。",
+ "presets.ocr_language": "認識言語",
+ "presets.ocr_language_hint": "文字や単語を正確に認識するために使用する言語。",
+ "presets.ocr_language.auto": "自動 / システムのデフォルト",
+ "presets.ocr_engine_required": "Tesseractが必要です。設定 → エンジンからインストールしてください",
+ "presets.ocr_disabled_linux": "Linuxには組み込みOCRがありません。有効にするには設定 → エンジンでTesseractをインストールしてください。",
"presets.group.media": "動画・音声",
"presets.media_quality": "品質",
"presets.media_quality_hint": "変換先が元のコーデックに対応している場合(iPhoneのMOV → MP4など)、動画を再エンコードせずに形式だけを変えます。\n画質を保ち、HDRも維持したまま数秒で完了します。\n「小さめ」はファイルサイズを減らすため、常に再エンコードします。",
@@ -652,6 +661,11 @@
"engines.guidance.linux.libreoffice": "パッケージマネージャーでインストール(Ubuntu/Debian:sudo apt install libreoffice)してから、「再確認」を選んでください。\nHWPを開くには、JavaとH2Orestart拡張機能も必要です。",
"engines.guidance.linux.ghostscript": "パッケージマネージャーでインストール(Ubuntu/Debian:sudo apt install ghostscript)してから、「再確認」を選んでください。",
"engines.guidance.windows": "このPC向けの{0}はFileeからダウンロードできません。\n「手元のプログラムを使う…」を選び、インストール済みのものを指定してください。",
+ "engines.package.tesseract.name": "Tesseract(多言語テキスト認識)",
+ "engines.package.tesseract.description": "デーヴァナーガリー(ネパール語/ヒンディー語)、アラビア語、希少言語を含む100以上の言語のテキストを認識します。\nmacOSではApple Visionが英語、CJK、ヨーロッパ言語をネイティブにサポートしています。",
+ "engines.guidance.macos.tesseract": "HomebrewでTesseractをインストール(brew install tesseract tesseract-lang)してから「再確認」を選択してください。\nまたは「手持ちのコピーを使う…」を選択してbinフォルダーを指定してください。",
+ "engines.guidance.linux.tesseract": "パッケージマネージャーでTesseractをインストール(Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all)してから「再確認」を選択してください。\nまたは「手持ちのコピーを使う…」を選択してプログラムフォルダーを指定してください。",
+ "engines.guidance.windows.tesseract": "UB-Mannheim(github.com/UB-Mannheim/tesseract)からTesseractをインストールし、「再確認」を選択するかインストールフォルダーを指定してください。",
"engines.use_own_app_folder": "{0}.appを含むフォルダーを選択してください",
"engines.progress_detail": "{0}/{1} · {2}/秒 · {3}",
"engines.eta.estimating": "残り時間を計算中…",
diff --git a/src/Filee.App/Assets/i18n/ko.json b/src/Filee.App/Assets/i18n/ko.json
index 115bfc7e..379ada5e 100644
--- a/src/Filee.App/Assets/i18n/ko.json
+++ b/src/Filee.App/Assets/i18n/ko.json
@@ -147,6 +147,7 @@
"engine.ghostscript.description": "EPS, 포스트스크립트, 예전 일러스트레이터 파일을 PDF와 이미지로, PDF를 EPS·PS로.",
"engine.calibre.description": "드문 전자책 형식(LIT, LRF, CHM, PDB, RB…)과 킨들용 출력(MOBI, AZW3).\nHWPX writer보다 위에 두면 전자책 PDF도 만들어요(긴 책은 훨씬 빨라요).",
"engine.libreoffice.description": "DOC, PPT, Pages, Keynote, Numbers, 워드퍼펙트, ClarisWorks, Publisher, Visio 같은 오래되었거나 드문 형식을 읽어요.\nDOC·XLS·PPT·ODT·ODP, 오픈도큐먼트 서식 파일, EMF로 저장도 해요.",
+ "engine.image-ocr.description": "스캔된 PDF 및 이미지에서 텍스트를 추출하여 편집 가능한 문서(Word, 텍스트, HWPX)로 변환합니다.",
"profile.images": "이미지",
"profile.pdf": "PDF",
@@ -534,6 +535,14 @@
"presets.leave_out_pictures": "그림 빼기 (텍스트만)",
"presets.leave_out_pictures_hint": "끄면: 그림을 마크다운 파일 옆 \"이름_files\" 폴더에 저장하고 파일에서 연결해요.\n켜면: 폴더 없이 텍스트만 남겨요.",
"presets.target_folder": "폴더 (압축 풀기)",
+ "presets.group.ocr": "광학 문자 인식 (OCR)",
+ "presets.ocr_enabled": "스캔 및 이미지에서 텍스트 인식",
+ "presets.ocr_hint": "스캔된 페이지와 이미지에서 편집 가능한 텍스트를 추출하여 Word, 텍스트 및 문서로 변환합니다.",
+ "presets.ocr_language": "인식 언어",
+ "presets.ocr_language_hint": "문자와 단어를 정확하게 인식하는 데 사용되는 언어입니다.",
+ "presets.ocr_language.auto": "자동 / 시스템 기본값",
+ "presets.ocr_engine_required": "Tesseract가 필요합니다. 설정 → 엔진에서 설치해 주세요",
+ "presets.ocr_disabled_linux": "Linux에는 내장 OCR이 없습니다. 활성화하려면 설정 → 엔진에서 Tesseract를 설치하세요.",
"presets.group.media": "동영상·오디오",
"presets.media_quality": "품질",
"presets.media_quality_hint": "새 형식에 코덱을 그대로 담을 수 있으면(아이폰 MOV → MP4 등) 다시 인코딩하지 않아요.\n화질과 HDR이 그대로고 몇 초면 끝나요.\n'작게'는 용량을 줄이려고 항상 다시 인코딩해요.",
@@ -669,6 +678,11 @@
"engines.guidance.linux.libreoffice": "패키지 관리자로 설치(Ubuntu/Debian: sudo apt install libreoffice)한 뒤 “다시 확인”을 누르세요.\nHWP를 열려면 Java와 H2Orestart 확장도 필요해요.",
"engines.guidance.linux.ghostscript": "패키지 관리자로 설치(Ubuntu/Debian: sudo apt install ghostscript)한 뒤 “다시 확인”을 누르세요.",
"engines.guidance.windows": "이 PC용 {0} 다운로드는 없어요.\n“내 프로그램 쓰기…”로 설치된 프로그램을 선택해 주세요.",
+ "engines.package.tesseract.name": "Tesseract (다국어 텍스트 인식)",
+ "engines.package.tesseract.description": "데바나가리(네팔어/힌디어), 아랍어 및 희귀 문자를 포함한 100개 이상의 언어로 텍스트를 인식합니다.\nmacOS에서는 Apple Vision이 내장되어 영어, CJK 및 유럽어는 추가 설치 없이 동작합니다.",
+ "engines.guidance.macos.tesseract": "Homebrew로 Tesseract를 설치(brew install tesseract tesseract-lang)한 뒤 “다시 확인”을 선택하세요.\n또는 “내 사본 사용…”을 선택하고 bin 폴더를 지정하세요.",
+ "engines.guidance.linux.tesseract": "패키지 관리자로 Tesseract를 설치(Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all)한 뒤 “다시 확인”을 선택하세요.\n또는 “내 사본 사용…”을 선택하고 프로그램 폴더를 지정하세요.",
+ "engines.guidance.windows.tesseract": "UB-Mannheim(github.com/UB-Mannheim/tesseract)에서 Tesseract를 설치한 뒤 “다시 확인”을 선택하거나 설치 폴더를 지정하세요.",
"engines.use_own_app_folder": "{0}.app이 들어 있는 폴더 선택",
"engines.progress_detail": "{0} / {1} · {2}/s · {3}",
"engines.eta.estimating": "남은 시간 계산 중…",
diff --git a/src/Filee.App/Assets/i18n/pl.json b/src/Filee.App/Assets/i18n/pl.json
index 2e2da36b..15e82d6b 100644
--- a/src/Filee.App/Assets/i18n/pl.json
+++ b/src/Filee.App/Assets/i18n/pl.json
@@ -147,6 +147,7 @@
"engine.ghostscript.description": "EPS, PostScript i starsze pliki Illustratora do PDF i obrazów; PDF do EPS i PS.",
"engine.calibre.description": "Rzadkie formaty e-booków (LIT, LRF, CHM, PDB, RB…) i zapis dla Kindle (MOBI, AZW3).\nUmieszczony nad HWPX writer tworzy też PDF-y e-booków (znacznie szybciej przy długich książkach).",
"engine.libreoffice.description": "Odczytuje starsze i rzadsze formaty: DOC, PPT, Pages, Keynote, Numbers, WordPerfect, ClarisWorks, Publisher, Visio…\nZapisuje też jako DOC, XLS, PPT, ODT, ODP, szablony OpenDocument i EMF.",
+ "engine.image-ocr.description": "Wyodrębnia tekst ze skanowanych plików PDF i obrazów do formatu edytowalnych dokumentów (Word, tekst, HWPX).",
"profile.images": "Obrazy",
"profile.pdf": "PDF",
@@ -534,6 +535,14 @@
"presets.leave_out_pictures": "Pomiń obrazy (tylko tekst)",
"presets.leave_out_pictures_hint": "Wyłączone: obrazy trafiają do folderu „nazwa_files” obok pliku Markdown, który do nich linkuje.\nWłączone: tylko tekst, bez folderu.",
"presets.target_folder": "Folder (rozpakuj archiwa)",
+ "presets.group.ocr": "Optyczne rozpoznawanie znaków (OCR)",
+ "presets.ocr_enabled": "Rozpoznaj tekst w skanach i obrazach",
+ "presets.ocr_hint": "Wyodrębnia edytowalny tekst ze skanowanych stron i obrazów do formatu Word, tekstu i dokumentów.",
+ "presets.ocr_language": "Język rozpoznawania",
+ "presets.ocr_language_hint": "Język używany do dokładnego rozpoznawania znaków i słów.",
+ "presets.ocr_language.auto": "Automatycznie / Domyślny systemu",
+ "presets.ocr_engine_required": "Wymaga Tesseract. Zainstaluj w Ustawienia → Silniki",
+ "presets.ocr_disabled_linux": "System Linux nie ma wbudowanego OCR. Zainstaluj Tesseract w Ustawienia → Silniki, aby włączyć.",
"presets.group.media": "Wideo i audio",
"presets.media_quality": "Jakość",
"presets.media_quality_hint": "Gdy nowy format może przyjąć kodeki bez zmian (MOV z iPhone’a → MP4), wideo jest tylko przepakowywane.\nBez utraty jakości, z zachowaniem HDR, w kilka sekund.\nMała zawsze koduje ponownie, by zmniejszyć plik.",
@@ -669,6 +678,11 @@
"engines.guidance.linux.libreoffice": "Zainstaluj menedżerem pakietów (Ubuntu/Debian: sudo apt install libreoffice), potem wybierz „Sprawdź ponownie”.\nOtwieranie HWP wymaga też Javy i rozszerzenia H2Orestart.",
"engines.guidance.linux.ghostscript": "Zainstaluj menedżerem pakietów (Ubuntu/Debian: sudo apt install ghostscript), potem wybierz „Sprawdź ponownie”.",
"engines.guidance.windows": "Filee nie ma do pobrania {0} dla tego komputera.\nWybierz „Użyj mojej kopii…” i wskaż zainstalowaną kopię.",
+ "engines.package.tesseract.name": "Tesseract (wielojęzyczne rozpoznawanie tekstu)",
+ "engines.package.tesseract.description": "Rozpoznaje tekst w ponad 100 językach, w tym dewanagari (nepalski/hindi), arabskim i rzadkich pismach.\nNa systemie macOS Apple Vision działa natywnie dla języka angielskiego, CJK i języków europejskich bez tego pakietu.",
+ "engines.guidance.macos.tesseract": "Zainstaluj Tesseract za pomocą Homebrew (brew install tesseract tesseract-lang), a następnie wybierz „Sprawdź ponownie”.\nLub wybierz „Użyj mojej kopii…” i wskaż folder bin.",
+ "engines.guidance.linux.tesseract": "Zainstaluj Tesseract za pomocą menedżera pakietów (Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all), a następnie wybierz „Sprawdź ponownie”.\nLub wybierz „Użyj mojej kopii…” i wskaż folder programu.",
+ "engines.guidance.windows.tesseract": "Zainstaluj Tesseract z UB-Mannheim (github.com/UB-Mannheim/tesseract), a następnie wybierz „Sprawdź ponownie” lub wskaż folder instalacji.",
"engines.use_own_app_folder": "Wybierz folder zawierający {0}.app",
"engines.progress_detail": "{0} z {1} · {2}/s · {3}",
"engines.eta.estimating": "szacowanie pozostałego czasu…",
diff --git a/src/Filee.App/Assets/i18n/pt-BR.json b/src/Filee.App/Assets/i18n/pt-BR.json
index 0b7cd7cd..ad32a285 100644
--- a/src/Filee.App/Assets/i18n/pt-BR.json
+++ b/src/Filee.App/Assets/i18n/pt-BR.json
@@ -147,6 +147,7 @@
"engine.ghostscript.description": "EPS, PostScript e arquivos antigos do Illustrator para PDF e imagens; PDF para EPS e PS.",
"engine.calibre.description": "Formatos raros de e-book (LIT, LRF, CHM, PDB, RB…) e saída Kindle (MOBI, AZW3).\nAcima do gravador de HWPX, também gera os PDFs de e-books (bem mais rápido em livros longos).",
"engine.libreoffice.description": "Lê formatos antigos e menos comuns: DOC, PPT, Pages, Keynote, Numbers, WordPerfect, ClarisWorks, Publisher, Visio…\nTambém salva como DOC, XLS, PPT, ODT, ODP, modelos OpenDocument e EMF.",
+ "engine.image-ocr.description": "Extrai texto de PDFs digitalizados e imagens para documentos editáveis (Word, texto, HWPX).",
"profile.images": "Imagens",
"profile.pdf": "PDF",
@@ -534,6 +535,14 @@
"presets.leave_out_pictures": "Deixar as imagens de fora (só texto)",
"presets.leave_out_pictures_hint": "Desligado: as imagens são salvas numa pasta “nome_files” ao lado do arquivo Markdown e referenciadas nele.\nLigado: só o texto, sem pasta.",
"presets.target_folder": "Pasta (extrair arquivos compactados)",
+ "presets.group.ocr": "Reconhecimento Óptico de Caracteres (OCR)",
+ "presets.ocr_enabled": "Reconhecer texto em digitalizações e imagens",
+ "presets.ocr_hint": "Extrai texto editável de páginas digitalizadas e imagens para Word, texto e documentos.",
+ "presets.ocr_language": "Idioma de reconhecimento",
+ "presets.ocr_language_hint": "Idioma usado para reconhecer caracteres e palavras com precisão.",
+ "presets.ocr_language.auto": "Automático / Padrão do sistema",
+ "presets.ocr_engine_required": "Requer Tesseract. Instale em Configurações → Motores",
+ "presets.ocr_disabled_linux": "O Linux não possui OCR integrado. Instale o Tesseract em Configurações → Motores para ativar.",
"presets.group.media": "Vídeo e áudio",
"presets.media_quality": "Qualidade",
"presets.media_quality_hint": "Quando o novo formato pode guardar os codecs como estão (MOV do iPhone → MP4), o vídeo só é reempacotado.\nSem perda de qualidade, HDR mantido, pronto em segundos.\nPequeno sempre recodifica para deixar o arquivo menor.",
@@ -669,6 +678,11 @@
"engines.guidance.linux.libreoffice": "Instale com o gerenciador de pacotes (Ubuntu/Debian: sudo apt install libreoffice) e escolha “Verificar de novo”.\nAbrir HWP também precisa de Java e da extensão H2Orestart.",
"engines.guidance.linux.ghostscript": "Instale com o gerenciador de pacotes (Ubuntu/Debian: sudo apt install ghostscript) e escolha “Verificar de novo”.",
"engines.guidance.windows": "O Filee não tem download de {0} para este PC.\nEscolha “Usar minha cópia…” e selecione uma cópia instalada.",
+ "engines.package.tesseract.name": "Tesseract (reconhecimento de texto multilíngue)",
+ "engines.package.tesseract.description": "Reconhece texto em mais de 100 idiomas, incluindo devanágari (nepalês/híndi), árabe e alfabetos raros.\nNo macOS, o Apple Vision funciona nativamente para inglês, CJK e idiomas europeus sem ele.",
+ "engines.guidance.macos.tesseract": "Instale o Tesseract com o Homebrew (brew install tesseract tesseract-lang) e selecione “Verificar novamente”.\nOu escolha “Usar minha cópia…” e selecione a pasta bin.",
+ "engines.guidance.linux.tesseract": "Instale o Tesseract com o gerenciador de pacotes (Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all) e selecione “Verificar novamente”.\nOu escolha “Usar minha cópia…” e selecione a pasta do programa.",
+ "engines.guidance.windows.tesseract": "Instale o Tesseract do UB-Mannheim (github.com/UB-Mannheim/tesseract) e selecione “Verificar novamente” ou selecione sua pasta de instalação.",
"engines.use_own_app_folder": "Escolha a pasta que contém {0}.app",
"engines.progress_detail": "{0} de {1} · {2}/s · {3}",
"engines.eta.estimating": "estimando o tempo restante…",
diff --git a/src/Filee.App/Assets/i18n/ru.json b/src/Filee.App/Assets/i18n/ru.json
index 3eb526c7..a12e691e 100644
--- a/src/Filee.App/Assets/i18n/ru.json
+++ b/src/Filee.App/Assets/i18n/ru.json
@@ -147,6 +147,7 @@
"engine.ghostscript.description": "EPS, PostScript и старые файлы Illustrator — в PDF и картинки; PDF — в EPS и PS.",
"engine.calibre.description": "Редкие форматы электронных книг (LIT, LRF, CHM, PDB, RB…) и вывод для Kindle (MOBI, AZW3).\nЕсли стоит выше HWPX writer, он же делает PDF из электронных книг (для длинных книг намного быстрее).",
"engine.libreoffice.description": "Читает старые и редкие форматы: DOC, PPT, Pages, Keynote, Numbers, WordPerfect, ClarisWorks, Publisher, Visio…\nА ещё сохраняет в DOC, XLS, PPT, ODT, ODP, шаблоны OpenDocument и EMF.",
+ "engine.image-ocr.description": "Извлекает текст из отсканированных PDF и изображений в редактируемые документы (Word, текст, HWPX).",
"profile.images": "Изображения",
"profile.pdf": "PDF",
@@ -534,6 +535,14 @@
"presets.leave_out_pictures": "Без картинок (только текст)",
"presets.leave_out_pictures_hint": "Выключено: картинки сохраняются в папку «имя_files» рядом с файлом Markdown и связаны с ним ссылками.\nВключено: только текст, без папки.",
"presets.target_folder": "Папка (распаковка архивов)",
+ "presets.group.ocr": "Оптическое распознавание символов (OCR)",
+ "presets.ocr_enabled": "Распознавать текст в сканах и изображениях",
+ "presets.ocr_hint": "Извлекает редактируемый текст из отсканированных страниц и изображений в Word, текст и документы.",
+ "presets.ocr_language": "Язык распознавания",
+ "presets.ocr_language_hint": "Язык для точного распознавания символов и слов.",
+ "presets.ocr_language.auto": "Автоматически / По умолчанию",
+ "presets.ocr_engine_required": "Требуется Tesseract. Установите его в Настройки → Движки",
+ "presets.ocr_disabled_linux": "В Linux нет встроенного OCR. Установите Tesseract в Настройки → Движки, чтобы включить.",
"presets.group.media": "Видео и аудио",
"presets.media_quality": "Качество",
"presets.media_quality_hint": "Если новый формат может принять кодеки как есть (MOV с iPhone → MP4), видео только переупаковывается.\nБез потери качества, HDR сохраняется, за секунды.\n«Компактное» всегда перекодирует, чтобы уменьшить файл.",
@@ -669,6 +678,11 @@
"engines.guidance.linux.libreoffice": "Установите менеджером пакетов (Ubuntu/Debian: sudo apt install libreoffice), затем выберите «Проверить снова».\nДля открытия HWP также нужны Java и расширение H2Orestart.",
"engines.guidance.linux.ghostscript": "Установите менеджером пакетов (Ubuntu/Debian: sudo apt install ghostscript), затем выберите «Проверить снова».",
"engines.guidance.windows": "Filee не может скачать {0} для этого компьютера.\nВыберите «Использовать свою копию…» и укажите установленную копию.",
+ "engines.package.tesseract.name": "Tesseract (многоязычное распознавание текста)",
+ "engines.package.tesseract.description": "Распознает текст на более чем 100 языках, включая деванагари (непальский/хинди), арабский и редкие письменности.\nВ macOS Apple Vision работает нативно для английского, CJK и европейских языков без него.",
+ "engines.guidance.macos.tesseract": "Установите Tesseract с помощью Homebrew (brew install tesseract tesseract-lang), затем выберите «Проверить снова».\nИли выберите «Использовать мою копию…» и укажите папку bin.",
+ "engines.guidance.linux.tesseract": "Установите Tesseract с помощью пакетного менеджера (Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all), затем выберите «Проверить снова».\nИли выберите «Использовать мою копию…» и укажите папку программы.",
+ "engines.guidance.windows.tesseract": "Установите Tesseract из UB-Mannheim (github.com/UB-Mannheim/tesseract), затем выберите «Проверить снова» или укажите папку установки.",
"engines.use_own_app_folder": "Выберите папку, в которой лежит {0}.app",
"engines.progress_detail": "{0} из {1} · {2}/с · {3}",
"engines.eta.estimating": "оцениваем, сколько осталось…",
diff --git a/src/Filee.App/Assets/i18n/zh-CN.json b/src/Filee.App/Assets/i18n/zh-CN.json
index 6247a2b0..e3342628 100644
--- a/src/Filee.App/Assets/i18n/zh-CN.json
+++ b/src/Filee.App/Assets/i18n/zh-CN.json
@@ -147,6 +147,7 @@
"engine.ghostscript.description": "EPS、PostScript 和旧版 Illustrator 文件转为 PDF 和图片;PDF 转为 EPS 和 PS。",
"engine.calibre.description": "少见的电子书格式(LIT、LRF、CHM、PDB、RB…)以及 Kindle 输出(MOBI、AZW3)。\n排在 HWPX writer 之上时,也由它把电子书转为 PDF(长书快得多)。",
"engine.libreoffice.description": "读取较旧或少见的格式,如 DOC、PPT、Pages、Keynote、Numbers、WordPerfect、ClarisWorks、Publisher、Visio。\n也能保存为 DOC、XLS、PPT、ODT、ODP、OpenDocument 模板或 EMF。",
+ "engine.image-ocr.description": "从扫描 PDF 和图片中提取文字并转换为可编辑文档(Word、文本、HWPX)。",
"profile.images": "图片",
"profile.pdf": "PDF",
@@ -534,6 +535,14 @@
"presets.leave_out_pictures": "不保留图片(仅文本)",
"presets.leave_out_pictures_hint": "关闭时,图片保存在 Markdown 文件旁的“名称_files”文件夹中,并从文件中链接。\n开启时,只保留文字,不生成文件夹。",
"presets.target_folder": "文件夹(解压缩)",
+ "presets.group.ocr": "光学字符识别 (OCR)",
+ "presets.ocr_enabled": "识别扫描件和图像中的文本",
+ "presets.ocr_hint": "从扫描页面和图像中提取可编辑文本并转换为 Word、文本及文档。",
+ "presets.ocr_language": "识别语言",
+ "presets.ocr_language_hint": "用于准确识别字符和文字的语言。",
+ "presets.ocr_language.auto": "自动 / 系统默认",
+ "presets.ocr_engine_required": "需要 Tesseract。请在“设置 → 转换引擎”中安装",
+ "presets.ocr_disabled_linux": "Linux 没有内置 OCR。请在“设置 → 转换引擎”中安装 Tesseract 以启用。",
"presets.group.media": "视频和音频",
"presets.media_quality": "质量",
"presets.media_quality_hint": "新格式能直接容纳原编码时(如 iPhone MOV → MP4),只重新封装。\n画质无损、保留 HDR,几秒即可完成。\n“小”总会重新编码以减小文件。",
@@ -669,6 +678,11 @@
"engines.guidance.linux.libreoffice": "用软件包管理器安装(Ubuntu/Debian 下运行 sudo apt install libreoffice)后点“重新检查”。\n打开 HWP 还需要 Java 和 H2Orestart 扩展。",
"engines.guidance.linux.ghostscript": "用软件包管理器安装(Ubuntu/Debian 下运行 sudo apt install ghostscript)后点“重新检查”。",
"engines.guidance.windows": "没有适用于这台电脑的 {0} 下载。\n点“使用我的程序…”选择已安装的程序。",
+ "engines.package.tesseract.name": "Tesseract(多语言文字识别)",
+ "engines.package.tesseract.description": "支持梵文(尼泊尔语/印地语)、阿拉伯语等 100 多种语言的文字识别。\n在 macOS 上,系统内置 Apple Vision 可直接识别英语、中日韩及主要欧洲语言。",
+ "engines.guidance.macos.tesseract": "使用 Homebrew 安装 Tesseract(brew install tesseract tesseract-lang),然后选择“重新检查”。\n或选择“使用我的副本…”并指定其 bin 文件夹。",
+ "engines.guidance.linux.tesseract": "使用包管理器安装 Tesseract(Ubuntu/Debian: sudo apt install tesseract-ocr tesseract-ocr-all),然后选择“重新检查”。\n或选择“使用我的副本…”并指定其程序文件夹。",
+ "engines.guidance.windows.tesseract": "从 UB-Mannheim (github.com/UB-Mannheim/tesseract) 安装 Tesseract,然后选择“重新检查”或指定其安装文件夹。",
"engines.use_own_app_folder": "选择包含 {0}.app 的文件夹",
"engines.progress_detail": "{0} / {1} · {2}/s · {3}",
"engines.eta.estimating": "正在估算剩余时间…",
diff --git a/src/Filee.App/Services/AppHost.cs b/src/Filee.App/Services/AppHost.cs
index f06ce0dc..de8a4e87 100644
--- a/src/Filee.App/Services/AppHost.cs
+++ b/src/Filee.App/Services/AppHost.cs
@@ -4,6 +4,7 @@ using Filee.App.Services.Triggers;
using Filee.App.ViewModels;
using Filee.Core.Conversion;
using Filee.Core.Localization;
+using Filee.Core.Ocr;
using Filee.Core.Platform;
using Filee.Core.Settings;
using Filee.Core.Web;
@@ -48,6 +49,8 @@ public static class AppHost
services.AddSingleton<ElevationService>();
// Conversion engines
+ if (OperatingSystem.IsMacOS())
+ OcrEngineRegistry.Register(new MacOsVisionOcrEngine());
services.AddSingleton(_ => new EngineEnvironment(dataDir));
services.AddSingleton(sp => EngineRegistry.CreateAll(sp.GetRequiredService<EngineEnvironment>(), sp.GetRequiredService<ILoggerFactory>()));
services.AddSingleton(sp => new ConverterCatalog(sp.GetRequiredService<IReadOnlyList<IConverter>>()));
diff --git a/src/Filee.App/Services/EngineDownloadService.cs b/src/Filee.App/Services/EngineDownloadService.cs
index 2bd41056..14cd8cac 100644
--- a/src/Filee.App/Services/EngineDownloadService.cs
+++ b/src/Filee.App/Services/EngineDownloadService.cs
@@ -44,7 +44,9 @@ public sealed partial class EnginePackageState : ObservableObject
public string Sizes => EngineDownloads.CanDownload(Package)
? _loc.Format("engines.package.size", FormatBytes(EngineDownloads.DownloadSize(Package)), FormatBytes(Package.InstalledSize))
- : _loc["engines.package.system_required"];
+ : CanSystemInstall(Package)
+ ? _loc.Format("engines.package.size", FormatBytes(350_000_000), FormatBytes(Package.InstalledSize))
+ : _loc["engines.package.system_required"];
/// <summary>
/// How to get an engine Filee doesn't download on this system (FFmpeg and Ghostscript from Homebrew on macOS,
@@ -54,7 +56,7 @@ public sealed partial class EnginePackageState : ObservableObject
{
get
{
- if (EngineDownloads.CanDownload(Package))
+ if (EngineDownloads.CanDownload(Package) || CanSystemInstall(Package))
return null;
var system = OperatingSystem.IsMacOS() ? "macos" : OperatingSystem.IsLinux() ? "linux" : "windows";
var key = $"engines.guidance.{system}.{Package.Id}";
@@ -114,7 +116,7 @@ public sealed partial class EnginePackageState : ObservableObject
/// <summary>The engine is there: Filee's download or the user's own copy.</summary>
public bool IsInstalled => Status is EnginePackageStatus.Installed or EnginePackageStatus.OwnCopy;
- public bool CanRemove => Status == EnginePackageStatus.Installed;
+ public bool CanRemove => Status == EnginePackageStatus.Installed || (Package.Id == "tesseract" && Status == EnginePackageStatus.OwnCopy && FromSystem);
/// <summary>The user chose this copy ("Stop using it" applies).</summary>
public bool UsesOwnCopy => Status == EnginePackageStatus.OwnCopy && !FromSystem;
public bool ShowsCopyFolder => Status == EnginePackageStatus.OwnCopy;
@@ -122,8 +124,33 @@ public sealed partial class EnginePackageState : ObservableObject
|| (Status == EnginePackageStatus.OwnCopy && FromSystem);
public bool IsBusy => Status is EnginePackageStatus.Queued or EnginePackageStatus.Downloading or EnginePackageStatus.Unpacking;
/// <summary>Also when a copy on the PC is used: Filee's download is the tested version and wins once installed.</summary>
- public bool CanInstall => EngineDownloads.CanDownload(Package) && (Status is EnginePackageStatus.NotInstalled or EnginePackageStatus.Failed
- || (Status == EnginePackageStatus.OwnCopy && FromSystem));
+ public bool CanInstall => (EngineDownloads.CanDownload(Package) && (Status is EnginePackageStatus.NotInstalled or EnginePackageStatus.Failed || (Status == EnginePackageStatus.OwnCopy && FromSystem)))
+ || (CanSystemInstall(Package) && Status is EnginePackageStatus.NotInstalled or EnginePackageStatus.Failed);
+
+ /// <summary>
+ /// Whether the app can drive an in-place install of <paramref name="package"/> on this machine: Homebrew on
+ /// macOS, winget on Windows. Linux has no handler here (package managers vary too much to drive one
+ /// generically), so it keeps the same guidance-text + "Use my copy…" flow as every other engine.
+ /// </summary>
+ internal static bool CanSystemInstall(EnginePackage package) =>
+ package.Id == "tesseract" && (OperatingSystem.IsMacOS() ? LocateBrew() is not null : OperatingSystem.IsWindows());
+
+ /// <summary>Homebrew's own binary, at its two standard prefixes or on PATH; null when it isn't installed.</summary>
+ internal static string? LocateBrew()
+ {
+ if (File.Exists("/opt/homebrew/bin/brew")) return "/opt/homebrew/bin/brew";
+ if (File.Exists("/usr/local/bin/brew")) return "/usr/local/bin/brew";
+ var path = Environment.GetEnvironmentVariable("PATH");
+ if (path is not null)
+ {
+ foreach (var part in path.Split(':'))
+ {
+ var candidate = Path.Combine(part, "brew");
+ if (File.Exists(candidate)) return candidate;
+ }
+ }
+ return null;
+ }
public string StatusText => Status switch
{
@@ -258,32 +285,40 @@ public sealed class EngineDownloadService
await _oneAtATime.WaitAsync(cancellation.Token);
try
{
- // Progress<T> reports on the UI thread it was created on.
- var clock = Stopwatch.StartNew();
- var rate = new TransferRate();
- var progress = new Progress<EngineInstallProgress>(p =>
+ if (state.Package.Id == "tesseract" && !EngineDownloads.CanDownload(state.Package))
{
- if (!state.IsBusy)
- return;
- state.Status = p.Stage == EngineInstallStage.Downloading ? EnginePackageStatus.Downloading : EnginePackageStatus.Unpacking;
- state.Progress = p.Fraction * 100;
- if (p.Stage == EngineInstallStage.Downloading && p.BytesTotal > 0)
- {
- rate.Add(clock.Elapsed, p.BytesDone);
- state.ShowTransfer(p.BytesDone, p.BytesTotal, rate);
- }
- else
+ await InstallTesseractSystemAsync(state, cancellation.Token);
+ }
+ else
+ {
+ // Progress<T> reports on the UI thread it was created on.
+ var clock = Stopwatch.StartNew();
+ var rate = new TransferRate();
+ var progress = new Progress<EngineInstallProgress>(p =>
{
- state.ProgressDetail = "";
- }
- });
- await Task.Run(() => _installer.InstallAsync(state.Package, progress, cancellation.Token), cancellation.Token);
+ if (!state.IsBusy)
+ return;
+ state.Status = p.Stage == EngineInstallStage.Downloading ? EnginePackageStatus.Downloading : EnginePackageStatus.Unpacking;
+ state.Progress = p.Fraction * 100;
+ if (p.Stage == EngineInstallStage.Downloading && p.BytesTotal > 0)
+ {
+ rate.Add(clock.Elapsed, p.BytesDone);
+ state.ShowTransfer(p.BytesDone, p.BytesTotal, rate);
+ }
+ else
+ {
+ state.ProgressDetail = "";
+ }
+ });
+ await Task.Run(() => _installer.InstallAsync(state.Package, progress, cancellation.Token), cancellation.Token);
+ }
}
finally
{
_oneAtATime.Release();
}
- state.Status = EnginePackageStatus.Installed;
+ EngineEnvironment.ForgetSystemCopies();
+ state.Status = StatusOf(state);
_catalog.Refresh();
if (state.Package.ConverterIds.Contains("libreoffice"))
WarmUpInBackground(TimeSpan.Zero);
@@ -305,6 +340,142 @@ public sealed class EngineDownloadService
}
}
+ private static async Task InstallTesseractSystemAsync(EnginePackageState state, CancellationToken cancellationToken)
+ {
+ try
+ {
+ if (OperatingSystem.IsMacOS())
+ {
+ var brew = EnginePackageState.LocateBrew();
+ if (brew is null)
+ throw new InvalidOperationException("Homebrew was not found. Install Homebrew from https://brew.sh first.");
+
+ state.Status = EnginePackageStatus.Downloading;
+ state.Progress = 5;
+ state.ProgressDetail = "Starting Homebrew install...";
+
+ var psi = new ProcessStartInfo
+ {
+ FileName = brew,
+ ArgumentList = { "install", "tesseract", "tesseract-lang" },
+ RedirectStandardOutput = true,
+ RedirectStandardError = true,
+ UseShellExecute = false,
+ CreateNoWindow = true
+ };
+ var brewDir = Path.GetDirectoryName(brew);
+ var currentPath = Environment.GetEnvironmentVariable("PATH") ?? "";
+ if (brewDir is not null && !currentPath.Contains(brewDir))
+ psi.Environment["PATH"] = brewDir + ":" + currentPath;
+ psi.Environment["HOMEBREW_NO_AUTO_UPDATE"] = "1";
+ psi.Environment["HOMEBREW_NO_ENV_HINTS"] = "1";
+
+ using var process = new Process { StartInfo = psi };
+ process.OutputDataReceived += (_, e) =>
+ {
+ if (!string.IsNullOrWhiteSpace(e.Data))
+ {
+ Dispatcher.UIThread.Post(() =>
+ {
+ var line = e.Data.Trim();
+ state.ProgressDetail = line;
+ if (line.StartsWith("==> Downloading", StringComparison.OrdinalIgnoreCase))
+ {
+ state.Status = EnginePackageStatus.Downloading;
+ if (state.Progress < 50) state.Progress += 10;
+ }
+ else if (line.StartsWith("==> Pouring", StringComparison.OrdinalIgnoreCase) ||
+ line.StartsWith("==> Installing", StringComparison.OrdinalIgnoreCase))
+ {
+ state.Status = EnginePackageStatus.Unpacking;
+ if (state.Progress < 90) state.Progress += 10;
+ }
+ });
+ }
+ };
+ process.ErrorDataReceived += (_, e) =>
+ {
+ if (!string.IsNullOrWhiteSpace(e.Data))
+ {
+ Dispatcher.UIThread.Post(() =>
+ {
+ state.ProgressDetail = e.Data.Trim();
+ });
+ }
+ };
+
+ process.Start();
+ process.BeginOutputReadLine();
+ process.BeginErrorReadLine();
+
+ using (cancellationToken.Register(() =>
+ {
+ try { process.Kill(true); } catch { }
+ }))
+ {
+ await process.WaitForExitAsync(cancellationToken);
+ }
+
+ if (process.ExitCode != 0)
+ throw new InvalidOperationException($"Homebrew install failed with exit code {process.ExitCode}.");
+
+ state.Progress = 100;
+ state.ProgressDetail = "";
+ }
+ else if (OperatingSystem.IsWindows())
+ {
+ state.Status = EnginePackageStatus.Downloading;
+ state.Progress = 10;
+ state.ProgressDetail = "Installing Tesseract with winget...";
+
+ var psi = new ProcessStartInfo
+ {
+ FileName = "winget",
+ ArgumentList = { "install", "UB-Mannheim.TesseractOCR", "--accept-source-agreements", "--accept-package-agreements" },
+ RedirectStandardOutput = true,
+ RedirectStandardError = true,
+ UseShellExecute = false,
+ CreateNoWindow = true
+ };
+
+ using var process = new Process { StartInfo = psi };
+ process.OutputDataReceived += (_, e) =>
+ {
+ if (!string.IsNullOrWhiteSpace(e.Data))
+ {
+ Dispatcher.UIThread.Post(() =>
+ {
+ state.ProgressDetail = e.Data.Trim();
+ });
+ }
+ };
+
+ process.Start();
+ process.BeginOutputReadLine();
+
+ using (cancellationToken.Register(() =>
+ {
+ try { process.Kill(true); } catch { }
+ }))
+ {
+ await process.WaitForExitAsync(cancellationToken);
+ }
+
+ if (process.ExitCode != 0)
+ throw new InvalidOperationException($"winget install failed with exit code {process.ExitCode}.");
+ }
+ else
+ {
+ throw new PlatformNotSupportedException("Automatic installation is not supported on this platform. Please install tesseract using your package manager.");
+ }
+ }
+ finally
+ {
+ state.Cancellation = null;
+ state.ProgressDetail = "";
+ }
+ }
+
public void Cancel(EnginePackageState state) => state.Cancellation?.Cancel();
/// <summary>
@@ -346,8 +517,45 @@ public sealed class EngineDownloadService
return false;
try
{
+ if (state.Package.Id == "tesseract")
+ {
+ if (OperatingSystem.IsMacOS())
+ {
+ var brew = EnginePackageState.LocateBrew();
+ if (brew is not null)
+ {
+ var psi = new ProcessStartInfo
+ {
+ FileName = brew,
+ ArgumentList = { "uninstall", "tesseract", "tesseract-lang" },
+ UseShellExecute = false,
+ CreateNoWindow = true
+ };
+ var brewDir = Path.GetDirectoryName(brew);
+ var currentPath = Environment.GetEnvironmentVariable("PATH") ?? "";
+ if (brewDir is not null && !currentPath.Contains(brewDir))
+ psi.Environment["PATH"] = brewDir + ":" + currentPath;
+ psi.Environment["HOMEBREW_NO_AUTO_UPDATE"] = "1";
+ psi.Environment["HOMEBREW_NO_ENV_HINTS"] = "1";
+ using var process = Process.Start(psi);
+ process?.WaitForExit();
+ }
+ }
+ else if (OperatingSystem.IsWindows())
+ {
+ using var process = Process.Start(new ProcessStartInfo
+ {
+ FileName = "winget",
+ ArgumentList = { "uninstall", "UB-Mannheim.TesseractOCR" },
+ UseShellExecute = false,
+ CreateNoWindow = true
+ });
+ process?.WaitForExit();
+ }
+ EngineEnvironment.ForgetSystemCopies();
+ }
_installer.Uninstall(state.Package);
- state.Status = EnginePackageStatus.NotInstalled;
+ state.Status = StatusOf(state);
_catalog.Refresh();
return true;
}
diff --git a/src/Filee.App/ViewModels/PresetEditorViewModel.cs b/src/Filee.App/ViewModels/PresetEditorViewModel.cs
index d76f638b..43bc2b5f 100644
--- a/src/Filee.App/ViewModels/PresetEditorViewModel.cs
+++ b/src/Filee.App/ViewModels/PresetEditorViewModel.cs
@@ -6,6 +6,7 @@
using CommunityToolkit.Mvvm.ComponentModel;
using Filee.Core.Formats;
using Filee.Core.Localization;
+using Filee.Core.Ocr;
using Filee.Core.Presets;
namespace Filee.App.ViewModels;
@@ -97,6 +98,94 @@ public sealed partial class PresetEditorViewModel : ObservableObject
AnimationSpeeds = [.. ImageOptions.AnimationSpeeds.Append(_edit.Image.PlaybackSpeed).Distinct().Order()
.Select(s => new Choice<double>(s, s == 1 ? loc["presets.animation_speed.original"] : SpeedLabel(s)))];
+ OcrLanguages =
+ [
+ new Choice<string?>(null, loc["presets.ocr_language.auto"]),
+ // Popular / Primary Languages
+ new Choice<string?>("en", "English"),
+ new Choice<string?>("ne", "Nepali / नेपाली"),
+ new Choice<string?>("hi", "Hindi / हिन्दी"),
+ new Choice<string?>("zh-Hans", "Chinese Simplified / 简体中文"),
+ new Choice<string?>("zh-Hant", "Chinese Traditional / 繁體中文"),
+ new Choice<string?>("ja", "Japanese / 日本語"),
+ new Choice<string?>("ko", "Korean / 한국어"),
+ new Choice<string?>("es", "Spanish / Español"),
+ new Choice<string?>("fr", "French / Français"),
+ new Choice<string?>("de", "German / Deutsch"),
+ new Choice<string?>("it", "Italian / Italiano"),
+ new Choice<string?>("pt", "Portuguese / Português"),
+ new Choice<string?>("ru", "Russian / Русский"),
+ new Choice<string?>("ar", "Arabic / العربية"),
+ new Choice<string?>("th", "Thai / ไทย"),
+ new Choice<string?>("el", "Greek / Ελληνικά"),
+ new Choice<string?>("uk", "Ukrainian / Українська"),
+ new Choice<string?>("tr", "Turkish / Türkçe"),
+ new Choice<string?>("vi", "Vietnamese / Tiếng Việt"),
+ new Choice<string?>("id", "Indonesian / Bahasa Indonesia"),
+ new Choice<string?>("pl", "Polish / Polski"),
+ new Choice<string?>("nl", "Dutch / Nederlands"),
+ new Choice<string?>("sv", "Swedish / Svenska"),
+ new Choice<string?>("da", "Danish / Dansk"),
+ new Choice<string?>("fi", "Finnish / Suomi"),
+ new Choice<string?>("no", "Norwegian / Norsk"),
+ new Choice<string?>("cs", "Czech / Čeština"),
+ new Choice<string?>("sk", "Slovak / Slovenčina"),
+ new Choice<string?>("hu", "Hungarian / Magyar"),
+ new Choice<string?>("ro", "Romanian / Română"),
+ new Choice<string?>("bg", "Bulgarian / Български"),
+ new Choice<string?>("hr", "Croatian / Hrvatski"),
+ new Choice<string?>("sr-Cyrl", "Serbian (Cyrillic) / Српски"),
+ new Choice<string?>("sr-Latn", "Serbian (Latin) / Srpski"),
+ new Choice<string?>("sl", "Slovenian / Slovenščina"),
+ new Choice<string?>("et", "Estonian / Eesti"),
+ new Choice<string?>("lv", "Latvian / Latviešu"),
+ new Choice<string?>("lt", "Lithuanian / Lietuvių"),
+ new Choice<string?>("be", "Belarusian / Беларуская"),
+ // South Asian / Indic Languages
+ new Choice<string?>("sa", "Sanskrit / संस्कृतम्"),
+ new Choice<string?>("mr", "Marathi / मराठी"),
+ new Choice<string?>("ta", "Tamil / தமிழ்"),
+ new Choice<string?>("te", "Telugu / తెలుగు"),
+ new Choice<string?>("ur", "Urdu / اردو"),
+ new Choice<string?>("fa", "Persian / فارسی"),
+ new Choice<string?>("ps", "Pashto / پښتو"),
+ new Choice<string?>("sd", "Sindhi / سنڌي"),
+ new Choice<string?>("ug", "Uyghur / ئۇيغۇرچە"),
+ // European, Regional & World Languages
+ new Choice<string?>("af", "Afrikaans"),
+ new Choice<string?>("az", "Azerbaijani / Azərbaycan"),
+ new Choice<string?>("bs", "Bosnian / Bosanski"),
+ new Choice<string?>("ca", "Catalan / Català"),
+ new Choice<string?>("cy", "Welsh / Cymraeg"),
+ new Choice<string?>("eu", "Basque / Euskara"),
+ new Choice<string?>("ga", "Irish / Gaeilge"),
+ new Choice<string?>("gl", "Galician / Galego"),
+ new Choice<string?>("is", "Icelandic / Íslenska"),
+ new Choice<string?>("ka", "Georgian / ქართული"),
+ new Choice<string?>("ku", "Kurdish / Kurdî"),
+ new Choice<string?>("la", "Latin / Latina"),
+ new Choice<string?>("lb", "Luxembourgish / Lëtzebuergesch"),
+ new Choice<string?>("mi", "Maori / Te Reo Māori"),
+ new Choice<string?>("mk", "Macedonian / Македонски"),
+ new Choice<string?>("mn", "Mongolian / Монгол"),
+ new Choice<string?>("ms", "Malay / Bahasa Melayu"),
+ new Choice<string?>("mt", "Maltese / Malti"),
+ new Choice<string?>("oc", "Occitan / Lenga d'òc"),
+ new Choice<string?>("qu", "Quechua / Runasimi"),
+ new Choice<string?>("sq", "Albanian / Shqip"),
+ new Choice<string?>("sw", "Swahili / Kiswahili"),
+ new Choice<string?>("tl", "Tagalog / Filipino"),
+ new Choice<string?>("uz", "Uzbek / Oʻzbekcha"),
+ new Choice<string?>("kk", "Kazakh / Қазақша"),
+ new Choice<string?>("ky", "Kyrgyz / Кыргызча"),
+ new Choice<string?>("tg", "Tajik / Тоҷикӣ"),
+ new Choice<string?>("tt", "Tatar / Татарча"),
+ ];
+ // Vision and Tesseract each claim a real, limited set of languages (SupportedLanguages); the old PaddleOCR
+ // engine claimed over 100 it could never actually recognize (no bundled model, no install path). Only offer
+ // what a registered engine can really attempt, so picking a language never leads straight to "engine missing".
+ OcrLanguages = [.. OcrLanguages.Where(c => c.Value is null || OcrEngineRegistry.AnyEngineSupports(c.Value))];
+
_name = string.IsNullOrWhiteSpace(_edit.Name) ? loc.DisplayName(_edit) : _edit.Name;
_target = Targets.FirstOrDefault(t => t.Value == _edit.TargetFormat) ?? Targets[1];
_resize = ResizeModes.First(r => r.Value == _edit.Image.Resize);
@@ -163,6 +252,13 @@ public sealed partial class PresetEditorViewModel : ObservableObject
_pattern = _edit.Output.FileNamePattern;
_keepDates = _edit.Output.KeepDates;
_zipResult = _edit.Output.ZipResult;
+ _ocr = _edit.Document.Ocr;
+ _selectedOcrLanguage = OcrLanguages.FirstOrDefault(l =>
+ l.Value == _edit.Document.OcrLanguage ||
+ (l.Value is not null && _edit.Document.OcrLanguage is not null &&
+ (_edit.Document.OcrLanguage.StartsWith(l.Value, StringComparison.OrdinalIgnoreCase) ||
+ l.Value.StartsWith(_edit.Document.OcrLanguage, StringComparison.OrdinalIgnoreCase))))
+ ?? OcrLanguages[0];
}
public string PresetId => _original.Id;
@@ -193,6 +289,7 @@ public sealed partial class PresetEditorViewModel : ObservableObject
/// <summary>Playback speeds of animations, 0.25× to 4× (ImageOptions.AnimationSpeed).</summary>
public IReadOnlyList<Choice<double>> AnimationSpeeds { get; }
+ public IReadOnlyList<Choice<string?>> OcrLanguages { get; }
[ObservableProperty, NotifyPropertyChangedFor(nameof(NameError), nameof(IsValid))]
private string _name;
@@ -206,7 +303,8 @@ public sealed partial class PresetEditorViewModel : ObservableObject
nameof(AudioCodecs), nameof(AudioCodecChoice), nameof(ShowAudioCodec),
nameof(ShowArchive), nameof(ShowArchiveOptions), nameof(ShowExtractNote),
nameof(ShowSaveEveryFrame), nameof(ShowAnimationSpeed), nameof(ShowTrace), nameof(ShowExtractMotionPhoto),
- nameof(ShowLivePhotoChoice), nameof(LivePhotoKinds), nameof(LivePhotoChoice), nameof(IsLivePhoto), nameof(LivePhotoHint))]
+ nameof(ShowLivePhotoChoice), nameof(LivePhotoKinds), nameof(LivePhotoChoice), nameof(IsLivePhoto), nameof(LivePhotoHint),
+ nameof(ShowOcr), nameof(ShowOcrLanguage), nameof(ShowOcrMissingWarning), nameof(OcrDisabledOnLinux))]
private Choice<string> _target;
private LivePhotoKind _livePhotoValue;
@@ -393,6 +491,8 @@ public sealed partial class PresetEditorViewModel : ObservableObject
/// <summary>Markdown: the text only, no picture folder next to it (DocumentOptions.LeaveOutPictures).</summary>
[ObservableProperty] private bool _leaveOutPictures;
+ [ObservableProperty, NotifyPropertyChangedFor(nameof(ShowOcrLanguage)), NotifyPropertyChangedFor(nameof(ShowOcrMissingWarning))] private bool _ocr;
+ [ObservableProperty, NotifyPropertyChangedFor(nameof(HasOcrEngine)), NotifyPropertyChangedFor(nameof(ShowOcrMissingWarning))] private Choice<string?> _selectedOcrLanguage;
[ObservableProperty] private string _subfolderName;
/// <summary>The subfolder only for a file that becomes several (OutputRule.SubfolderOnlyForSeveralFiles).</summary>
[ObservableProperty] private bool _subfolderOnlyForSeveralFiles;
@@ -497,6 +597,17 @@ public sealed partial class PresetEditorViewModel : ObservableObject
public bool ShowDocument => TargetId == "pdf";
/// <summary>Markdown can't hold pictures: they go into a folder next to it, or are left out.</summary>
public bool ShowLeaveOutPictures => TargetId == "md";
+
+ /// <summary>OCR (Optical Character Recognition) for text recognition in scanned documents / images.</summary>
+ public bool ShowOcr => !IsTool && (TargetCategory is FormatCategory.Document or FormatCategory.Text or FormatCategory.Ebook
+ || TargetId is "docx" or "txt" or "md" or "hwpx" or "pdf" or "rtf" or "odt");
+ public bool ShowOcrLanguage => ShowOcr && Ocr;
+ public bool HasOcrEngine => OcrEngineRegistry.GetDefault(SelectedOcrLanguage?.Value) is not null;
+ public bool HasAnyOcrEngine => OcrEngineRegistry.All.Any(e => e.IsAvailable);
+ /// <summary>Linux has no built-in OCR (unlike Apple Vision on macOS), so Tesseract isn't just optional there: with
+ /// nothing installed, OCR can't run at all, and the toggle is disabled rather than left on to fail at conversion time.</summary>
+ public bool OcrDisabledOnLinux => ShowOcr && OperatingSystem.IsLinux() && !HasAnyOcrEngine;
+ public bool ShowOcrMissingWarning => ShowOcr && !OcrDisabledOnLinux && (!HasAnyOcrEngine || (Ocr && !HasOcrEngine));
public bool ShowMedia => ShowMediaQuality || ShowMaxHeight || ShowRemoveAudio || ShowAudioChannels || ShowVideoCodec || ShowAudioCodec;
/// <summary>Mono / stereo for every audio target but AMR, which is mono by definition.</summary>
public bool ShowAudioChannels => TargetIsAudio && TargetId != "amr";
@@ -598,6 +709,8 @@ public sealed partial class PresetEditorViewModel : ObservableObject
_original.Document.PdfA = PdfA;
// Only where the option is shown, as with SaveEveryFrame below.
_original.Document.LeaveOutPictures = ShowLeaveOutPictures && LeaveOutPictures;
+ _original.Document.Ocr = ShowOcr && Ocr;
+ _original.Document.OcrLanguage = ShowOcr && Ocr ? SelectedOcrLanguage.Value : null;
var media = _original.Media;
media.Quality = MediaQuality.Value;
diff --git a/src/Filee.App/Views/PresetEditorView.axaml b/src/Filee.App/Views/PresetEditorView.axaml
index 8c81df87..de37f45a 100644
--- a/src/Filee.App/Views/PresetEditorView.axaml
+++ b/src/Filee.App/Views/PresetEditorView.axaml
@@ -296,6 +296,38 @@
</StackPanel>
</Border>
+ <!-- Documents: Optical Character Recognition (OCR) for scanned pages and images -->
+ <Border x:Name="OcrCard" Classes="card" Padding="14" IsVisible="{Binding ShowOcr}">
+ <StackPanel Spacing="10">
+ <TextBlock Classes="h2" Text="{DynamicResource presets.group.ocr}" />
+ <StackPanel Spacing="2">
+ <CheckBox Content="{DynamicResource presets.ocr_enabled}" IsChecked="{Binding Ocr}" IsEnabled="{Binding !OcrDisabledOnLinux}" />
+ <TextBlock Classes="caption" TextWrapping="Wrap" Margin="28,0,0,0" Text="{DynamicResource presets.ocr_hint}" />
+ </StackPanel>
+ <Border Classes="badge" Margin="28,2,0,0" IsVisible="{Binding OcrDisabledOnLinux}" HorizontalAlignment="Left">
+ <StackPanel Orientation="Horizontal" Spacing="6">
+ <PathIcon Width="12" Height="12" Data="{StaticResource Icon.Alert}" Foreground="{DynamicResource WarningBrush}" />
+ <TextBlock Classes="caption" Foreground="{DynamicResource WarningBrush}" Text="{DynamicResource presets.ocr_disabled_linux}" />
+ </StackPanel>
+ </Border>
+ <Border Classes="badge" Margin="28,2,0,0" IsVisible="{Binding ShowOcrMissingWarning}" HorizontalAlignment="Left">
+ <StackPanel Orientation="Horizontal" Spacing="6">
+ <PathIcon Width="12" Height="12" Data="{StaticResource Icon.Alert}" Foreground="{DynamicResource WarningBrush}" />
+ <TextBlock Classes="caption" Foreground="{DynamicResource WarningBrush}" Text="{DynamicResource presets.ocr_engine_required}" />
+ </StackPanel>
+ </Border>
+ <StackPanel Spacing="2" Margin="28,4,0,0" IsVisible="{Binding ShowOcrLanguage}">
+ <Grid ColumnDefinitions="*,12,*">
+ <StackPanel>
+ <TextBlock Classes="label" Text="{DynamicResource presets.ocr_language}" />
+ <ComboBox ItemsSource="{Binding OcrLanguages}" SelectedItem="{Binding SelectedOcrLanguage}" HorizontalAlignment="Stretch" />
+ </StackPanel>
+ </Grid>
+ <TextBlock Classes="caption" TextWrapping="Wrap" Margin="0,4,0,0" Text="{DynamicResource presets.ocr_language_hint}" />
+ </StackPanel>
+ </StackPanel>
+ </Border>
+
<!-- PDF options (render settings also apply when a PDF is converted to images) -->
<Border Classes="card" Padding="14" IsVisible="{Binding ShowPdf}">
<StackPanel Spacing="10">
diff --git a/src/Filee.Cli/CliHost.cs b/src/Filee.Cli/CliHost.cs
index 14bb8b43..dae34da3 100644
--- a/src/Filee.Cli/CliHost.cs
+++ b/src/Filee.Cli/CliHost.cs
@@ -34,6 +34,8 @@ internal sealed class CliHost : IAsyncDisposable
EngineEnvironment.LibreOfficeForDocuments = store.Settings.LibreOfficeForDocuments;
EngineEnvironment.LibreOfficeForSheets = store.Settings.LibreOfficeForSheets;
var converters = EngineRegistry.CreateAll(new EngineEnvironment(store.Directory));
+ if (OperatingSystem.IsMacOS())
+ Filee.Core.Ocr.OcrEngineRegistry.Register(new Filee.Platform.MacOS.MacOsVisionOcrEngine());
var catalog = new ConverterCatalog(converters) { Priority = store.Settings.EnginePriority };
var queue = new JobQueue(catalog, EngineRegistry.FindPdfMerger(converters), combiner: EngineRegistry.FindFileCombiner(converters),
tiffMerger: EngineRegistry.FindTiffMerger(converters));
diff --git a/src/Filee.Cli/Filee.Cli.csproj b/src/Filee.Cli/Filee.Cli.csproj
index af728be3..1883ea6d 100644
--- a/src/Filee.Cli/Filee.Cli.csproj
+++ b/src/Filee.Cli/Filee.Cli.csproj
@@ -14,6 +14,7 @@
<ItemGroup>
<ProjectReference Include="..\Filee.Core\Filee.Core.csproj" />
<ProjectReference Include="..\Filee.Engines\Filee.Engines.csproj" />
+ <ProjectReference Include="..\Filee.Platform.MacOS\Filee.Platform.MacOS.csproj" Condition="$([MSBuild]::IsOSPlatform('OSX'))" />
</ItemGroup>
<!-- Error messages and built-in preset names, in English (the app's translation file). -->
diff --git a/src/Filee.Core/Ocr/IOcrEngine.cs b/src/Filee.Core/Ocr/IOcrEngine.cs
new file mode 100644
index 00000000..4f320abd
--- /dev/null
+++ b/src/Filee.Core/Ocr/IOcrEngine.cs
@@ -0,0 +1,30 @@
+// The contract every OCR engine implements (Apple Vision, Tesseract): recognize text and word boxes in an image,
+// so callers like PdfDocumentReader can place the result back onto the page without knowing which engine ran.
+
+namespace Filee.Core.Ocr;
+
+/// <summary>
+/// A service that extracts text and bounding boxes from images.
+/// </summary>
+public interface IOcrEngine
+{
+ /// <summary>Unique engine identifier (e.g. "vision", "tesseract").</summary>
+ string Id { get; }
+
+ /// <summary>Display name (e.g. "Apple Vision", "Tesseract OCR").</summary>
+ string DisplayName { get; }
+
+ /// <summary>Whether this engine is available and ready on the current machine.</summary>
+ bool IsAvailable { get; }
+
+ /// <summary>Supported languages (BCP-47 codes, or empty if generic/all).</summary>
+ IReadOnlyList<string> SupportedLanguages { get; }
+
+ /// <summary>
+ /// Recognizes text in an image file.
+ /// </summary>
+ /// <param name="imagePath">Path to an image file (PNG, JPG, TIFF, etc.).</param>
+ /// <param name="language">Optional language hint or code.</param>
+ /// <param name="cancellationToken">Cancellation token.</param>
+ Task<OcrPageResult> RecognizeAsync(string imagePath, string? language = null, CancellationToken cancellationToken = default);
+}
diff --git a/src/Filee.Core/Ocr/OcrEngineRegistry.cs b/src/Filee.Core/Ocr/OcrEngineRegistry.cs
new file mode 100644
index 00000000..504bdf7d
--- /dev/null
+++ b/src/Filee.Core/Ocr/OcrEngineRegistry.cs
@@ -0,0 +1,135 @@
+// Where callers (PdfDocumentReader, ImageOcrConverter, the preset editor) find an OCR engine without depending on
+// Filee.Platform.MacOS or Filee.Engines directly: Tesseract registers itself through EngineRegistry.CreateAll on
+// every platform, Apple Vision is discovered lazily and by reflection since it lives in a macOS-only assembly.
+
+namespace Filee.Core.Ocr;
+
+/// <summary>
+/// Registry of available OCR engines.
+/// </summary>
+public static class OcrEngineRegistry
+{
+ private static readonly List<IOcrEngine> RegisteredEngines = [];
+ private static readonly Lock Sync = new();
+
+ public static void Register(IOcrEngine engine)
+ {
+ ArgumentNullException.ThrowIfNull(engine);
+ lock (Sync)
+ {
+ if (!RegisteredEngines.Any(e => e.Id.Equals(engine.Id, StringComparison.OrdinalIgnoreCase)))
+ {
+ RegisteredEngines.Add(engine);
+ }
+ }
+ }
+
+ public static IReadOnlyList<IOcrEngine> All
+ {
+ get
+ {
+ lock (Sync)
+ {
+ DiscoverNativeEngine();
+ return [.. RegisteredEngines];
+ }
+ }
+ }
+
+ /// <summary>
+ /// The engine to use. With a specific <paramref name="language"/>, only an engine whose
+ /// <see cref="IOcrEngine.SupportedLanguages"/> actually covers it is returned — never an engine that would
+ /// silently misread a script it doesn't know (e.g. Apple Vision guessing at Devanagari), even if that engine is
+ /// otherwise the platform's native/preferred one. With no language ("Automatic"), the platform's native engine
+ /// (Apple Vision on macOS: zero install, GPU/NPU accelerated) is preferred, since the document's script isn't
+ /// known ahead of time; picking a non-Latin-only language from the preset editor is what routes to Tesseract.
+ /// </summary>
+ public static IOcrEngine? GetDefault(string? language = null)
+ {
+ lock (Sync)
+ {
+ DiscoverNativeEngine();
+
+ if (!string.IsNullOrWhiteSpace(language))
+ return RegisteredEngines.FirstOrDefault(e => e.IsAvailable && Supports(e, language));
+
+ var native = RegisteredEngines.FirstOrDefault(e => e.Id.Equals("vision", StringComparison.OrdinalIgnoreCase) && e.IsAvailable);
+ return native ?? RegisteredEngines.FirstOrDefault(e => e.IsAvailable);
+ }
+ }
+
+ /// <summary>
+ /// The exact code to pass to <paramref name="engine"/>'s <see cref="IOcrEngine.RecognizeAsync"/> for Filee's
+ /// <paramref name="language"/> code: Apple Vision needs its own full BCP-47 entry ("en" → "en-US"), engines that
+ /// map Filee's codes themselves (Tesseract) get the code unchanged.
+ /// </summary>
+ public static string? ResolveLanguage(IOcrEngine? engine, string? language)
+ {
+ if (engine is null || string.IsNullOrWhiteSpace(language))
+ return language;
+ return engine.SupportedLanguages.FirstOrDefault(l =>
+ l.Equals(language, StringComparison.OrdinalIgnoreCase)
+ || l.StartsWith(language + "-", StringComparison.OrdinalIgnoreCase)) ?? language;
+ }
+
+ /// <summary>
+ /// Whether any registered engine (installed or not) claims to support <paramref name="language"/> — for the
+ /// preset editor's language picker, so it never offers a language no engine could ever attempt. Use
+ /// <see cref="GetDefault"/> to find the engine that can actually run it right now.
+ /// </summary>
+ public static bool AnyEngineSupports(string language)
+ {
+ lock (Sync)
+ {
+ DiscoverNativeEngine();
+ return RegisteredEngines.Any(e => Supports(e, language));
+ }
+ }
+
+ private static bool Supports(IOcrEngine engine, string language) =>
+ engine.SupportedLanguages.Count == 0
+ || engine.SupportedLanguages.Any(l => l.Equals(language, StringComparison.OrdinalIgnoreCase)
+ || l.StartsWith(language + "-", StringComparison.OrdinalIgnoreCase));
+
+ /// <summary>Registers the platform's native engine (Apple Vision) the first time it's needed.</summary>
+ private static void DiscoverNativeEngine()
+ {
+ if (!OperatingSystem.IsMacOS() || RegisteredEngines.Any(e => e.Id.Equals("vision", StringComparison.OrdinalIgnoreCase)))
+ return;
+ try
+ {
+ var asm = AppDomain.CurrentDomain.GetAssemblies().FirstOrDefault(a => a.GetName().Name == "Filee.Platform.MacOS");
+ if (asm is null)
+ {
+ try
+ {
+ asm = System.Runtime.Loader.AssemblyLoadContext.Default.LoadFromAssemblyName(new System.Reflection.AssemblyName("Filee.Platform.MacOS"));
+ }
+ catch
+ {
+ var baseDir = AppContext.BaseDirectory;
+ var path = System.IO.Path.Combine(baseDir, "Filee.Platform.MacOS.dll");
+ if (System.IO.File.Exists(path))
+ asm = System.Runtime.Loader.AssemblyLoadContext.Default.LoadFromAssemblyPath(path);
+ }
+ }
+
+ var macEngineType = asm?.GetType("Filee.Platform.MacOS.MacOsVisionOcrEngine")
+ ?? Type.GetType("Filee.Platform.MacOS.MacOsVisionOcrEngine, Filee.Platform.MacOS");
+ if (macEngineType is not null && Activator.CreateInstance(macEngineType) is IOcrEngine macEngine && macEngine.IsAvailable)
+ RegisteredEngines.Add(macEngine);
+ }
+ catch
+ {
+ // Ignore discovery failure
+ }
+ }
+
+ public static IOcrEngine? GetById(string id)
+ {
+ lock (Sync)
+ {
+ return RegisteredEngines.FirstOrDefault(e => e.Id.Equals(id, StringComparison.OrdinalIgnoreCase));
+ }
+ }
+}
diff --git a/src/Filee.Core/Ocr/OcrResult.cs b/src/Filee.Core/Ocr/OcrResult.cs
new file mode 100644
index 00000000..2d60ac93
--- /dev/null
+++ b/src/Filee.Core/Ocr/OcrResult.cs
@@ -0,0 +1,39 @@
+// The result of an IOcrEngine.RecognizeAsync call: text and boxes in image pixel coordinates, grouped into words
+// and lines so a caller can either read the plain text or reconstruct layout (PdfDocumentReader scales these boxes
+// back onto the PDF page, one shared baseline per line rather than per word).
+
+namespace Filee.Core.Ocr;
+
+/// <summary>
+/// A bounding box in image pixel coordinates (Left/Top to Right/Bottom).
+/// </summary>
+public readonly record struct OcrRect(double Left, double Top, double Right, double Bottom)
+{
+ public double Width => Math.Max(0, Right - Left);
+ public double Height => Math.Max(0, Bottom - Top);
+}
+
+/// <summary>
+/// A recognized word with its text, bounding box, and recognition confidence.
+/// </summary>
+public sealed record OcrWord(string Text, OcrRect Box, float Confidence = 1.0f);
+
+/// <summary>
+/// A recognized line composed of words.
+/// </summary>
+public sealed record OcrLine(string Text, IReadOnlyList<OcrWord> Words, OcrRect Box);
+
+/// <summary>
+/// The complete OCR result of an image or page.
+/// </summary>
+public sealed record OcrPageResult(
+ string Text,
+ IReadOnlyList<OcrLine> Lines,
+ IReadOnlyList<OcrWord> Words,
+ int ImageWidth,
+ int ImageHeight,
+ string? Language = null)
+{
+ public static OcrPageResult Empty(int width = 0, int height = 0) =>
+ new("", [], [], width, height);
+}
diff --git a/src/Filee.Core/Presets/Preset.cs b/src/Filee.Core/Presets/Preset.cs
index ed631510..eada26a5 100644
--- a/src/Filee.Core/Presets/Preset.cs
+++ b/src/Filee.Core/Presets/Preset.cs
@@ -423,6 +423,16 @@ public sealed class DocumentOptions
/// </summary>
public bool LeaveOutPictures { get; set; }
+ /// <summary>
+ /// Optical Character Recognition for scanned pages and images.
+ /// </summary>
+ public bool Ocr { get; set; } = true;
+
+ /// <summary>
+ /// Optical Character Recognition language code (e.g. "en-US", "ko-KR"). Null uses the system default.
+ /// </summary>
+ public string? OcrLanguage { get; set; }
+
public DocumentOptions Clone() => (DocumentOptions)MemberwiseClone();
}
diff --git a/src/Filee.Engines/EngineRegistry.cs b/src/Filee.Engines/EngineRegistry.cs
index b9c0f010..facebb67 100644
--- a/src/Filee.Engines/EngineRegistry.cs
+++ b/src/Filee.Engines/EngineRegistry.cs
@@ -1,6 +1,7 @@
// THE place where conversion engines are registered. Adding an engine = one line in CreateAll.
using Filee.Core.Conversion;
+using Filee.Core.Ocr;
using Filee.Engines.Archives;
using Filee.Engines.Cad;
using Filee.Engines.Data;
@@ -15,6 +16,7 @@ using Filee.Engines.Infrastructure;
using Filee.Engines.Magick;
using Filee.Engines.Media;
using Filee.Engines.Models3D;
+using Filee.Engines.Ocr;
using Filee.Engines.Office;
using Filee.Engines.Pdf;
using Filee.Engines.Subtitles;
@@ -37,38 +39,44 @@ public static class EngineRegistry
/// The app's log; FFmpeg writes which hardware encoders work and when one failed, LibreOffice when a file needed
/// its container repacked.
/// </param>
- public static IReadOnlyList<IConverter> CreateAll(EngineEnvironment env, ILoggerFactory? loggers = null) =>
- [
- new MagickImageConverter(),
- new PdfSharpConverter(),
- new PdfiumConverter(),
- new RhwpConverter(),
- new UnhwpConverter(),
- new MarkdownConverter(),
- new SpreadsheetConverter(),
- new OoxmlConverter(),
- new HwpxConverter(),
- new ArchiveConverter(),
- new VectorConverter(),
- new IcnsConverter(),
- new FontConverter(),
- new DocxWriterConverter(),
- new PdfTextConverter(),
- new PdfSlidesConverter(),
- new EmlConverter(),
- new DataConverter(),
- new SubtitleConverter(),
- new WicImageConverter(),
- new FfmpegConverter(loggers?.CreateLogger<FfmpegConverter>()),
- new CadConverter(),
- new ModelConverter(),
- new GeoConverter(),
- new EbookConverter(),
- new PandocConverter(),
- new GhostscriptConverter(),
- new CalibreConverter(),
- new LibreOfficeConverter(env, loggers?.CreateLogger<LibreOfficeConverter>()),
- ];
+ public static IReadOnlyList<IConverter> CreateAll(EngineEnvironment env, ILoggerFactory? loggers = null)
+ {
+ OcrEngineRegistry.Register(new TesseractOcrEngine());
+
+ return
+ [
+ new MagickImageConverter(),
+ new PdfSharpConverter(),
+ new PdfiumConverter(),
+ new RhwpConverter(),
+ new UnhwpConverter(),
+ new MarkdownConverter(),
+ new SpreadsheetConverter(),
+ new OoxmlConverter(),
+ new HwpxConverter(),
+ new ArchiveConverter(),
+ new VectorConverter(),
+ new IcnsConverter(),
+ new FontConverter(),
+ new DocxWriterConverter(),
+ new ImageOcrConverter(),
+ new PdfTextConverter(),
+ new PdfSlidesConverter(),
+ new EmlConverter(),
+ new DataConverter(),
+ new SubtitleConverter(),
+ new WicImageConverter(),
+ new FfmpegConverter(loggers?.CreateLogger<FfmpegConverter>()),
+ new CadConverter(),
+ new ModelConverter(),
+ new GeoConverter(),
+ new EbookConverter(),
+ new PandocConverter(),
+ new GhostscriptConverter(),
+ new CalibreConverter(),
+ new LibreOfficeConverter(env, loggers?.CreateLogger<LibreOfficeConverter>()),
+ ];
+ }
/// <summary>Finds the PDF merger among the converters.</summary>
public static IPdfMerger? FindPdfMerger(IEnumerable<IConverter> converters) =>
diff --git a/src/Filee.Engines/Filee.Engines.csproj b/src/Filee.Engines/Filee.Engines.csproj
index ca144acd..1c3d7420 100644
--- a/src/Filee.Engines/Filee.Engines.csproj
+++ b/src/Filee.Engines/Filee.Engines.csproj
@@ -39,6 +39,7 @@
<InternalsVisibleTo Include="Filee" />
<InternalsVisibleTo Include="Filee.Engines.Tests" />
<InternalsVisibleTo Include="Filee.App.Tests" />
+ <InternalsVisibleTo Include="Filee.Platform.Tests" />
</ItemGroup>
<ItemGroup>
diff --git a/src/Filee.Engines/Hwp/Hwpx/DocumentReaders.cs b/src/Filee.Engines/Hwp/Hwpx/DocumentReaders.cs
index b98e78d0..2920b655 100644
--- a/src/Filee.Engines/Hwp/Hwpx/DocumentReaders.cs
+++ b/src/Filee.Engines/Hwp/Hwpx/DocumentReaders.cs
@@ -4,6 +4,7 @@
using System.Text;
using System.Text.Json.Nodes;
+using Filee.Core.Presets;
using Filee.Engines.Ebooks;
using Filee.Engines.Email;
using Filee.Engines.Hwp.Hwpx.Docx;
@@ -105,9 +106,15 @@ internal static class DocumentReaders
/// <summary>Reads <paramref name="path"/> (of format <paramref name="format"/>) into the document model.</summary>
/// <param name="workDirectory">Scratch folder of the job: extracted pictures go into a new folder inside it.</param>
- public static async Task<HDocument> ReadAsync(string path, string format, string workDirectory, CancellationToken cancellationToken)
+ public static async Task<HDocument> ReadAsync(string path, string format, string workDirectory, CancellationToken cancellationToken, Preset? preset = null)
{
var media = Path.Combine(workDirectory, $"media-{Guid.NewGuid():N}");
+ if (format.Equals("pdf", StringComparison.OrdinalIgnoreCase))
+ {
+ var enableOcr = preset?.Document.Ocr ?? true;
+ var ocrLang = preset?.Document.OcrLanguage;
+ return await Task.Run(() => PdfDocumentReader.Read(path, media, null, ocrLang, enableOcr), cancellationToken);
+ }
if (BuiltIn.TryGetValue(format, out var reader))
return await Task.Run(() => reader(path, media), cancellationToken);
if (ThroughPandoc.TryGetValue(format, out var pandocReader))
diff --git a/src/Filee.Engines/Hwp/Hwpx/HwpxConverter.cs b/src/Filee.Engines/Hwp/Hwpx/HwpxConverter.cs
index bf6f570e..fdaa1d19 100644
--- a/src/Filee.Engines/Hwp/Hwpx/HwpxConverter.cs
+++ b/src/Filee.Engines/Hwp/Hwpx/HwpxConverter.cs
@@ -45,7 +45,7 @@ public sealed class HwpxConverter : IConverter
public async Task<IReadOnlyList<string>> ConvertAsync(ConversionStep step, IProgress<double>? progress, CancellationToken cancellationToken)
{
progress?.Report(0.1);
- var document = await DocumentReaders.ReadAsync(step.InputPath, step.From, step.WorkDirectory, cancellationToken);
+ var document = await DocumentReaders.ReadAsync(step.InputPath, step.From, step.WorkDirectory, cancellationToken, step.Preset);
progress?.Report(0.6);
var output = step.Output.Allocate("hwpx");
diff --git a/src/Filee.Engines/Infrastructure/EngineDownloads.cs b/src/Filee.Engines/Infrastructure/EngineDownloads.cs
index 7ae3e302..4a150684 100644
--- a/src/Filee.Engines/Infrastructure/EngineDownloads.cs
+++ b/src/Filee.Engines/Infrastructure/EngineDownloads.cs
@@ -44,12 +44,17 @@ public static class EngineDownloads
new("ghostscript", OperatingSystem.IsWindows() ? ["ghostscript", "vcruntime"] : ["ghostscript"], 31_000_000, ["ghostscript"]),
new("ffmpeg", ["ffmpeg"], 272_000_000, ["ffmpeg"]),
new("calibre", ["calibre"], 663_000_000, ["calibre"]),
+ // No Filee-hosted download on any platform yet (Tesseract's Linux/macOS builds are dynamically linked
+ // against leptonica/libpng/etc., not xcopy-deployable the way FFmpeg's static builds are): install it with
+ // the OS's own package manager or installer, then "Use my copy…", same as Ghostscript on macOS and Linux.
+ new("tesseract", [], 30_000_000, ["image-ocr"]),
];
/// <summary>Total download size of a package in bytes.</summary>
public static long DownloadSize(EnginePackage package) => package.Components.Sum(c => Components.GetValueOrDefault(c)?.Size ?? 0);
- public static bool CanDownload(EnginePackage package) => package.Components.Count > 0 && package.Components.All(Components.ContainsKey);
+ public static bool CanDownload(EnginePackage package) =>
+ package.Components.Count > 0 && package.Components.All(Components.ContainsKey);
public static string? UnavailableReason(EnginePackage package)
{
@@ -61,6 +66,7 @@ public static class EngineDownloads
"ffmpeg" => "Install FFmpeg with Homebrew (brew install ffmpeg), then check again or select its bin folder.",
"ghostscript" => "Install Ghostscript with Homebrew (brew install ghostscript), then check again or select its bin folder.",
"calibre" => "The calibre Filee downloads needs macOS 14 or later. Install calibre 7.26 from calibre-ebook.com and select its .app folder.",
+ "tesseract" => "Install Tesseract with Homebrew (brew install tesseract tesseract-lang for every language), then check again or select its bin folder. Apple Vision already recognizes English, CJK and a few other languages without it.",
_ => $"No Filee download is available for {RuntimeIdentifier}. Install {package.Id} for this Mac and select its program or .app folder.",
};
if (OperatingSystem.IsLinux())
@@ -68,8 +74,11 @@ public static class EngineDownloads
{
"libreoffice" => "Install LibreOffice using your distribution's package manager (Ubuntu/Debian: sudo apt install libreoffice), then check again. HWP import through LibreOffice also needs Java and the H2Orestart extension.",
"ghostscript" => "Install Ghostscript using your distribution's package manager (Ubuntu/Debian: sudo apt install ghostscript), then check again.",
+ "tesseract" => "Install Tesseract using your distribution's package manager (Ubuntu/Debian: sudo apt install tesseract-ocr, plus tesseract-ocr-<lang> for each language, e.g. tesseract-ocr-nep), then check again. Linux has no built-in OCR, so this is required for OCR to work at all.",
_ => $"No Filee download is available for {RuntimeIdentifier}. Install {package.Id} using your distribution's package manager and select its program folder.",
};
+ if (package.Id == "tesseract")
+ return "Install Tesseract from https://github.com/UB-Mannheim/tesseract (ticking the languages you need), then check again or select its install folder.";
return $"No Filee download is available for {RuntimeIdentifier}. Select a compatible installed copy.";
}
diff --git a/src/Filee.Engines/Infrastructure/EngineEnvironment.cs b/src/Filee.Engines/Infrastructure/EngineEnvironment.cs
index a94979e6..12a9da9e 100644
--- a/src/Filee.Engines/Infrastructure/EngineEnvironment.cs
+++ b/src/Filee.Engines/Infrastructure/EngineEnvironment.cs
@@ -84,6 +84,7 @@ public sealed class EngineEnvironment(string dataDirectory)
["calibre"] = [windows ? "ebook-convert.exe" : "ebook-convert"],
["rhwp"] = [windows ? "rhwp.exe" : "rhwp"],
["7zip"] = [windows ? "7z.exe" : "7zz"],
+ ["tesseract"] = [windows ? "tesseract.exe" : "tesseract"],
};
public static string ProgramName(string name) => OperatingSystem.IsWindows() ? name
@@ -217,6 +218,8 @@ public sealed class EngineEnvironment(string dataDirectory)
}
if (id == "calibre")
folders.AddRange(["/Applications/calibre.app", Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.UserProfile), "Applications", "calibre.app"), "/opt/calibre"]);
+ if (id == "tesseract")
+ folders.AddRange(["/opt/homebrew/opt/tesseract/bin", "/usr/local/opt/tesseract/bin"]);
}
foreach (var root in (OperatingSystem.IsWindows()
? new[] { Environment.SpecialFolder.ProgramFiles, Environment.SpecialFolder.ProgramFilesX86 }
@@ -231,6 +234,10 @@ public sealed class EngineEnvironment(string dataDirectory)
case "calibre":
folders.Add(Path.Combine(root, "Calibre2"));
break;
+ case "tesseract":
+ // The official UB-Mannheim installer's default location.
+ folders.Add(Path.Combine(root, "Tesseract-OCR"));
+ break;
case "ghostscript":
// C:\Program Files\gs\gs10.04.0\bin, newest first
try
diff --git a/src/Filee.Engines/Infrastructure/EngineInstaller.cs b/src/Filee.Engines/Infrastructure/EngineInstaller.cs
index 12fe5fac..ec2cdc70 100644
--- a/src/Filee.Engines/Infrastructure/EngineInstaller.cs
+++ b/src/Filee.Engines/Infrastructure/EngineInstaller.cs
@@ -53,11 +53,11 @@ public sealed class EngineInstaller
}
/// <summary>True when every component is installed (any version).</summary>
- public bool IsInstalled(EnginePackage package) => package.Components.All(c => _components.ContainsKey(c) && Marker(c) is not null);
+ public bool IsInstalled(EnginePackage package) => package.Components.Count > 0 && package.Components.All(c => _components.ContainsKey(c) && Marker(c) is not null);
/// <summary>True when every component is installed in the version engines.json pins.</summary>
public bool IsUpToDate(EnginePackage package) =>
- package.Components.All(c => _components.TryGetValue(c, out var component)
+ package.Components.Count > 0 && package.Components.All(c => _components.TryGetValue(c, out var component)
&& string.Equals(Marker(c), component.Sha256, StringComparison.OrdinalIgnoreCase));
/// <summary>Downloads and unpacks the components that are missing or outdated.</summary>
diff --git a/src/Filee.Engines/Ocr/ImageOcrConverter.cs b/src/Filee.Engines/Ocr/ImageOcrConverter.cs
new file mode 100644
index 00000000..64e2d217
--- /dev/null
+++ b/src/Filee.Engines/Ocr/ImageOcrConverter.cs
@@ -0,0 +1,96 @@
+// Image → TXT, DOCX and Markdown via whichever OCR engine OcrEngineRegistry resolves (Apple Vision on macOS,
+// Tesseract everywhere else). Allows dropping scans, photos or screenshots directly and extracting editable text.
+
+using System.Text;
+using Filee.Core.Conversion;
+using Filee.Core.Ocr;
+using Filee.Engines.Hwp.Hwpx;
+using Filee.Engines.Hwp.Hwpx.Docx;
+using Filee.Engines.Infrastructure;
+
+namespace Filee.Engines.Ocr;
+
+/// <summary>
+/// Converts raster images directly to text documents using Optical Character Recognition.
+/// </summary>
+public sealed class ImageOcrConverter : IConverter
+{
+ private static readonly string[] ImageFormats = ["png", "jpg", "jpeg", "webp", "tiff", "bmp"];
+
+ public string Id => "image-ocr";
+ public string DisplayName => "Image OCR (built-in)";
+ public int MaxParallelism => 0;
+
+ public IReadOnlyList<ConversionEdge> Edges { get; } =
+ ImageFormats.SelectMany(from => new ConversionEdge[]
+ {
+ new(from, "txt", 15, DirectOnly: true),
+ new(from, "docx", 15, DirectOnly: true),
+ new(from, "md", 15, DirectOnly: true),
+ }).ToList();
+
+ public EngineStatus GetStatus()
+ {
+ var engine = OcrEngineRegistry.GetDefault();
+ if (engine is { IsAvailable: true })
+ return EngineStatus.Available($"Image → TXT, DOCX, Markdown ({engine.DisplayName})", EngineVersions.BuiltIn);
+
+ return EngineStatus.Unavailable("engine.reason.not_installed", "No supported OCR engine found on this system.");
+ }
+
+ public async Task<IReadOnlyList<string>> ConvertAsync(ConversionStep step, IProgress<double>? progress, CancellationToken cancellationToken)
+ {
+ if (!step.Preset.Document.Ocr)
+ throw new InvalidOperationException("OCR is disabled in preset settings.");
+
+ var language = step.Preset.Document.OcrLanguage;
+ var engine = OcrEngineRegistry.GetDefault(language);
+ if (engine is null || !engine.IsAvailable)
+ throw new InvalidOperationException(string.IsNullOrWhiteSpace(language)
+ ? "No OCR engine is installed. Install Tesseract in Settings → Engines."
+ : $"No installed OCR engine recognizes '{language}'. Install Tesseract with that language's trained data in Settings → Engines.");
+
+ progress?.Report(0.1);
+ var result = await engine.RecognizeAsync(step.InputPath, OcrEngineRegistry.ResolveLanguage(engine, language), cancellationToken);
+ cancellationToken.ThrowIfCancellationRequested();
+ progress?.Report(0.7);
+
+ if (string.IsNullOrWhiteSpace(result.Text))
+ throw new InvalidOperationException("No text could be recognized in the image.");
+
+ var output = step.Output.Allocate(step.To);
+ if (output is null)
+ return [];
+
+ switch (step.To)
+ {
+ case "txt":
+ await File.WriteAllTextAsync(output, result.Text, new UTF8Encoding(encoderShouldEmitUTF8Identifier: true), cancellationToken);
+ break;
+
+ case "md":
+ var mdContent = string.Join("\n\n", result.Lines.Select(l => l.Text));
+ await File.WriteAllTextAsync(output, mdContent, new UTF8Encoding(encoderShouldEmitUTF8Identifier: false), cancellationToken);
+ break;
+
+ case "docx":
+ var doc = new HDocument();
+ var section = new HSection();
+ foreach (var line in result.Lines)
+ {
+ var p = new HParagraph();
+ p.Inlines.Add(new HText(line.Text, default));
+ section.Blocks.Add(p);
+ }
+ doc.Sections.Add(section);
+ await Task.Run(() => DocxWriter.Write(doc, output), cancellationToken);
+ break;
+
+ default:
+ throw new NotSupportedException($"Target format {step.To} is not supported by Image OCR.");
+ }
+
+ progress?.Report(1.0);
+ return [output];
+ }
+}
diff --git a/src/Filee.Engines/Ocr/TesseractOcrEngine.cs b/src/Filee.Engines/Ocr/TesseractOcrEngine.cs
new file mode 100644
index 00000000..2baea8ce
--- /dev/null
+++ b/src/Filee.Engines/Ocr/TesseractOcrEngine.cs
@@ -0,0 +1,243 @@
+// Tesseract OCR engine: a single self-contained binary plus pinned .traineddata files, found the same way as every
+// other optional engine (OwnCopies, Filee's own copy, then one already on the PC). The real cross-platform, no-Python
+// fallback for platforms without a native OCR API (Windows, Linux) and for scripts Apple Vision doesn't recognize.
+
+using System.Globalization;
+using Filee.Core.Ocr;
+using Filee.Engines.Infrastructure;
+
+namespace Filee.Engines.Ocr;
+
+/// <summary>
+/// Optical Character Recognition engine powered by Tesseract (Apache-2.0), run as a subprocess. Covers every
+/// platform and, with the matching .traineddata installed, scripts Apple Vision does not (Devanagari, Cyrillic
+/// beyond Russian/Ukrainian, Thai, Greek, Arabic script, ...).
+/// </summary>
+public sealed class TesseractOcrEngine : IOcrEngine
+{
+ private static readonly TimeSpan Timeout = TimeSpan.FromMinutes(2);
+
+ public string Id => "tesseract";
+ public string DisplayName => "Tesseract OCR";
+
+ public bool IsAvailable => LocateExecutable() is not null;
+
+ /// <summary>
+ /// Filee's language codes this engine can be asked to recognize, limited to what Tesseract's own trained-data
+ /// set actually ships (tessdata_fast/best): asking for a code outside this list always fails, so the UI and
+ /// <see cref="OcrEngineRegistry"/> only offer what can really work, instead of PaddleOCR's old claimed-but-never
+ /// bundled 100+ languages.
+ /// </summary>
+ public IReadOnlyList<string> SupportedLanguages => [.. LanguageMap.Keys];
+
+ /// <summary>Filee language code (the OCR language picker, BCP-47-ish) → Tesseract's own 3-letter code.</summary>
+ private static readonly IReadOnlyDictionary<string, string> LanguageMap = new Dictionary<string, string>(StringComparer.OrdinalIgnoreCase)
+ {
+ ["en"] = "eng",
+ ["ne"] = "nep",
+ ["hi"] = "hin",
+ ["zh-Hans"] = "chi_sim",
+ ["zh-Hant"] = "chi_tra",
+ ["ja"] = "jpn",
+ ["ko"] = "kor",
+ ["es"] = "spa",
+ ["fr"] = "fra",
+ ["de"] = "deu",
+ ["it"] = "ita",
+ ["pt"] = "por",
+ ["ru"] = "rus",
+ ["ar"] = "ara",
+ ["th"] = "tha",
+ ["el"] = "ell",
+ ["uk"] = "ukr",
+ ["tr"] = "tur",
+ ["vi"] = "vie",
+ ["id"] = "ind",
+ ["pl"] = "pol",
+ ["nl"] = "nld",
+ ["sv"] = "swe",
+ ["da"] = "dan",
+ ["fi"] = "fin",
+ ["no"] = "nor",
+ ["cs"] = "ces",
+ ["sk"] = "slk",
+ ["hu"] = "hun",
+ ["ro"] = "ron",
+ ["bg"] = "bul",
+ ["hr"] = "hrv",
+ ["sr-Cyrl"] = "srp",
+ ["sr-Latn"] = "srp_latn",
+ ["sl"] = "slv",
+ ["et"] = "est",
+ ["lv"] = "lav",
+ ["lt"] = "lit",
+ ["be"] = "bel",
+ ["sa"] = "san",
+ ["mr"] = "mar",
+ ["ta"] = "tam",
+ ["te"] = "tel",
+ ["ur"] = "urd",
+ ["fa"] = "fas",
+ ["ps"] = "pus",
+ ["sd"] = "snd",
+ ["ug"] = "uig",
+ ["af"] = "afr",
+ ["az"] = "aze",
+ ["bs"] = "bos",
+ ["ca"] = "cat",
+ ["cy"] = "cym",
+ ["eu"] = "eus",
+ ["ga"] = "gle",
+ ["gl"] = "glg",
+ ["is"] = "isl",
+ ["ka"] = "kat",
+ ["ku"] = "kmr",
+ ["la"] = "lat",
+ ["lb"] = "ltz",
+ ["mi"] = "mri",
+ ["mk"] = "mkd",
+ ["mn"] = "mon",
+ ["ms"] = "msa",
+ ["mt"] = "mlt",
+ ["oc"] = "oci",
+ ["qu"] = "que",
+ ["sq"] = "sqi",
+ ["sw"] = "swa",
+ ["tl"] = "fil",
+ ["uz"] = "uzb",
+ ["kk"] = "kaz",
+ ["ky"] = "kir",
+ ["tg"] = "tgk",
+ ["tt"] = "tat",
+ };
+
+ /// <summary>Tesseract's own code for a Filee language code, or null when Tesseract has no matching trained data.</summary>
+ public static string? MapLanguage(string? language)
+ {
+ if (string.IsNullOrWhiteSpace(language))
+ return "eng";
+ return LanguageMap.TryGetValue(language.Trim(), out var code) ? code : null;
+ }
+
+ public static string? LocateExecutable() =>
+ EngineEnvironment.OwnProgram("tesseract", "tesseract.exe")
+ ?? EngineEnvironment.BundledProgram("tesseract", "tesseract.exe")
+ ?? EngineEnvironment.SystemProgram("tesseract", "tesseract.exe");
+
+ /// <summary>
+ /// The tessdata folder next to <paramref name="executable"/>: Homebrew, apt and the official Windows installer
+ /// all keep trained-data files at one of these relative locations. Null lets tesseract fall back to its own
+ /// compiled-in default (TESSDATA_PREFIX or the path it was built with), same as running it from a shell.
+ /// </summary>
+ public static string? LocateTessdataDir(string executable)
+ {
+ var bin = Path.GetDirectoryName(executable);
+ if (bin is null)
+ return null;
+ var root = Path.GetFullPath(Path.Combine(bin, ".."));
+ string[] candidates =
+ [
+ Path.Combine(bin, "tessdata"),
+ Path.Combine(root, "share", "tessdata"),
+ Path.Combine(root, "share", "tesseract-ocr", "5", "tessdata"),
+ Path.Combine(root, "share", "tesseract-ocr", "4.00", "tessdata"),
+ "/usr/share/tesseract-ocr/5/tessdata",
+ "/usr/share/tesseract-ocr/4.00/tessdata",
+ "/usr/share/tessdata",
+ ];
+ return candidates.FirstOrDefault(c => Directory.Exists(c) && Directory.EnumerateFiles(c, "*.traineddata").Any());
+ }
+
+ /// <summary>Whether <paramref name="language"/> has its .traineddata installed, not just a known mapping.</summary>
+ public bool HasLanguageData(string? language)
+ {
+ var exe = LocateExecutable();
+ var code = MapLanguage(language);
+ if (exe is null || code is null)
+ return false;
+ var dir = LocateTessdataDir(exe);
+ if (dir is null)
+ return true; // Can't check; let tesseract itself report a missing language if it turns out to be wrong.
+ return File.Exists(Path.Combine(dir, $"{code}.traineddata"));
+ }
+
+ public async Task<OcrPageResult> RecognizeAsync(string imagePath, string? language = null, CancellationToken cancellationToken = default)
+ {
+ if (!File.Exists(imagePath))
+ throw new FileNotFoundException("Image file not found for OCR.", imagePath);
+
+ var exe = LocateExecutable() ?? throw new InvalidOperationException("Tesseract is not installed.");
+ var code = MapLanguage(language)
+ ?? throw new InvalidOperationException($"Tesseract has no trained data for language '{language}'.");
+
+ var args = new List<string> { imagePath, "stdout", "-l", code };
+ if (LocateTessdataDir(exe) is { } tessdata)
+ args.AddRange(["--tessdata-dir", tessdata]);
+ args.Add("tsv");
+
+ var result = await ProcessRunner.RunAsync(exe, [.. args], Timeout, cancellationToken);
+ if (result.ExitCode != 0)
+ {
+ throw new InvalidOperationException(result.StandardError.Contains("Failed loading language", StringComparison.OrdinalIgnoreCase)
+ ? $"Tesseract has no trained data installed for '{code}'. Install the {language} language pack and try again."
+ : $"Tesseract could not read the image. {result.StandardError.Trim()}");
+ }
+
+ return ParseTsv(result.StandardOutput, code);
+ }
+
+ private static OcrPageResult ParseTsv(string tsv, string language)
+ {
+ var words = new List<OcrWord>();
+ var lineGroups = new List<(string Key, List<OcrWord> Words)>();
+ var imgWidth = 0;
+ var imgHeight = 0;
+
+ foreach (var row in tsv.Split('\n', StringSplitOptions.RemoveEmptyEntries))
+ {
+ var cols = row.Split('\t');
+ // level, page_num, block_num, par_num, line_num, word_num, left, top, width, height, conf, text
+ if (cols.Length < 12)
+ continue;
+ if (cols[0] == "1")
+ {
+ // Page-level row: left/top/width/height is the whole image, needed to scale word boxes back
+ // to the source page (PdfDocumentReader) even when the recognized text doesn't fill the image.
+ int.TryParse(cols[8], out imgWidth);
+ int.TryParse(cols[9], out imgHeight);
+ continue;
+ }
+ if (cols[0] != "5")
+ continue;
+ var text = cols[11].Trim();
+ if (text.Length == 0)
+ continue;
+ if (!double.TryParse(cols[6], NumberStyles.Float, CultureInfo.InvariantCulture, out var left)
+ || !double.TryParse(cols[7], NumberStyles.Float, CultureInfo.InvariantCulture, out var top)
+ || !double.TryParse(cols[8], NumberStyles.Float, CultureInfo.InvariantCulture, out var width)
+ || !double.TryParse(cols[9], NumberStyles.Float, CultureInfo.InvariantCulture, out var height))
+ continue;
+ float.TryParse(cols[10], NumberStyles.Float, CultureInfo.InvariantCulture, out var conf);
+
+ var word = new OcrWord(text, new OcrRect(left, top, left + width, top + height), Math.Max(0, conf) / 100f);
+ words.Add(word);
+
+ var lineKey = string.Join('.', cols[1], cols[2], cols[3], cols[4]);
+ var group = lineGroups.Count > 0 && lineGroups[^1].Key == lineKey ? lineGroups[^1] : (lineKey, []);
+ if (lineGroups.Count == 0 || lineGroups[^1].Key != lineKey)
+ lineGroups.Add(group);
+ group.Words.Add(word);
+ }
+
+ var lines = lineGroups.Select(g =>
+ {
+ var text = string.Join(' ', g.Words.Select(w => w.Text));
+ var box = new OcrRect(
+ g.Words.Min(w => w.Box.Left), g.Words.Min(w => w.Box.Top),
+ g.Words.Max(w => w.Box.Right), g.Words.Max(w => w.Box.Bottom));
+ return new OcrLine(text, g.Words, box);
+ }).ToList();
+
+ return new OcrPageResult(string.Join('\n', lines.Select(l => l.Text)), lines, words, imgWidth, imgHeight, language);
+ }
+}
diff --git a/src/Filee.Engines/Office/DocxWriterConverter.cs b/src/Filee.Engines/Office/DocxWriterConverter.cs
index 0b2deb23..9fc095ed 100644
--- a/src/Filee.Engines/Office/DocxWriterConverter.cs
+++ b/src/Filee.Engines/Office/DocxWriterConverter.cs
@@ -32,7 +32,7 @@ public sealed class DocxWriterConverter : IConverter
public async Task<IReadOnlyList<string>> ConvertAsync(ConversionStep step, IProgress<double>? progress, CancellationToken cancellationToken)
{
progress?.Report(0.1);
- var document = await DocumentReaders.ReadAsync(step.InputPath, step.From, step.WorkDirectory, cancellationToken);
+ var document = await DocumentReaders.ReadAsync(step.InputPath, step.From, step.WorkDirectory, cancellationToken, step.Preset);
progress?.Report(0.6);
var output = step.Output.Allocate("docx");
diff --git a/src/Filee.Engines/Pdf/PdfDocumentReader.cs b/src/Filee.Engines/Pdf/PdfDocumentReader.cs
index cb4a9058..bd67cde4 100644
--- a/src/Filee.Engines/Pdf/PdfDocumentReader.cs
+++ b/src/Filee.Engines/Pdf/PdfDocumentReader.cs
@@ -17,6 +17,7 @@
using System.Text;
using System.Text.RegularExpressions;
+using Filee.Core.Ocr;
using Filee.Engines.Hwp.Hwpx;
using Filee.Engines.Infrastructure;
using PDFtoImage;
@@ -102,23 +103,37 @@ internal sealed partial class PdfDocumentReader
private readonly string _path;
private readonly string _mediaFolder;
private readonly bool _textOnly;
+ private readonly IOcrEngine? _ocrEngine;
+ private readonly string? _ocrLanguage;
+ private readonly bool _enableOcr;
private byte[]? _pdfBytes;
private int _pictures;
- private PdfDocumentReader(string path, string mediaFolder, bool textOnly)
+ private PdfDocumentReader(string path, string mediaFolder, bool textOnly, IOcrEngine? ocrEngine = null, string? ocrLanguage = null, bool enableOcr = true)
{
_path = path;
_mediaFolder = mediaFolder;
_textOnly = textOnly;
+ _ocrEngine = ocrEngine;
+ _ocrLanguage = ocrLanguage;
+ _enableOcr = enableOcr;
}
/// <summary>Reads <paramref name="path"/> with pictures (extracted into <paramref name="mediaFolder"/>) and scans.</summary>
- public static HDocument Read(string path, string mediaFolder) => new PdfDocumentReader(path, mediaFolder, textOnly: false).ReadDocument();
+ public static HDocument Read(string path, string mediaFolder) =>
+ new PdfDocumentReader(path, mediaFolder, textOnly: false).ReadDocument();
+
+ public static HDocument Read(string path, string mediaFolder, IOcrEngine? ocrEngine, string? ocrLanguage = null, bool enableOcr = true) =>
+ new PdfDocumentReader(path, mediaFolder, textOnly: false, ocrEngine, ocrLanguage, enableOcr).ReadDocument();
/// <summary>
/// Reads only the text, including invisible text (the OCR layer of scanned PDFs); pictures and scans are left out.
/// </summary>
- public static HDocument ReadText(string path) => new PdfDocumentReader(path, "", textOnly: true).ReadDocument();
+ public static HDocument ReadText(string path) =>
+ new PdfDocumentReader(path, "", textOnly: true).ReadDocument();
+
+ public static HDocument ReadText(string path, IOcrEngine? ocrEngine, string? ocrLanguage = null, bool enableOcr = true) =>
+ new PdfDocumentReader(path, "", textOnly: true, ocrEngine, ocrLanguage, enableOcr).ReadDocument();
// ───────────────────────── Model of a page ─────────────────────────
@@ -237,7 +252,7 @@ internal sealed partial class PdfDocumentReader
new("The PDF is password-protected. Open it with its password and save a copy without one, then convert that copy.");
/// <summary>
- /// The text of each page in <paramref name="pages"/> (zero-based), read like <see cref="ReadText"/> (reading order,
+ /// The text of each page in <paramref name="pages"/> (zero-based), read like <see cref="ReadText(string)"/> (reading order,
/// invisible OCR text included) but page by page, with running heads and page numbers kept: "" for a page without
/// text. Paragraphs are separated by an empty line, lines end with CR LF.
/// </summary>
@@ -308,13 +323,83 @@ internal sealed partial class PdfDocumentReader
lead.StartBaseLine.Y - crop.Bottom, Math.Max(1, size), Format(lead, size), link));
}
+ if (words.Count == 0)
+ {
+ // First check if the page already had an invisible OCR layer that was skipped for formatting
+ foreach (var word in pageWords)
+ {
+ if (string.IsNullOrWhiteSpace(word.Text))
+ continue;
+ var letters = word.Letters.Where(l => !string.IsNullOrWhiteSpace(l.Value)).ToList();
+ if (letters.Count == 0)
+ continue;
+ var lead = string.IsNullOrWhiteSpace(word.Lead.Value) ? letters[0] : word.Lead;
+ var box = Shift(word.Box, crop);
+ var size = Median(letters.Select(LetterSize));
+ var center = new PdfPoint((box.Left + box.Right) / 2, (box.Top + box.Bottom) / 2);
+ var link = links.FirstOrDefault(l => Inside(l.Bounds, center)).Uri;
+ words.Add(new TextWord(CleanText(word.Text), box.Left, box.Right, box.Top, box.Bottom,
+ lead.StartBaseLine.Y - crop.Bottom, Math.Max(1, size), Format(lead, size), link));
+ }
+ }
+
if (words.Count == 0 || (rotated && !_textOnly))
{
- // A scan, a page of drawings, or a turned page: the rendered page keeps everything.
- content.Scanned = true;
- if (!_textOnly)
- content.ScanPath = RenderScan(page.Number - 1);
- return content;
+ if (_enableOcr)
+ {
+ var ocr = _ocrEngine ?? OcrEngineRegistry.GetDefault(_ocrLanguage);
+ if (ocr is not null && ocr.IsAvailable)
+ {
+ var scanPath = RenderScan(page.Number - 1);
+ try
+ {
+ var ocrResult = ocr.RecognizeAsync(scanPath, OcrEngineRegistry.ResolveLanguage(ocr, _ocrLanguage)).GetAwaiter().GetResult();
+ if (ocrResult.Words.Count > 0)
+ {
+ var scaleX = content.Width / (double)ocrResult.ImageWidth;
+ var scaleY = content.Height / (double)ocrResult.ImageHeight;
+ // One baseline/top/bottom per OCR-detected line, not per word: on a skewed or photographed
+ // scan each word's own pixel box drifts slightly in Y, which (once scaled to PDF points)
+ // is enough for Lines() below to split a single real text line into several interleaved
+ // reconstructed lines. Tesseract already groups words into lines; use that box.
+ foreach (var line in ocrResult.Lines)
+ {
+ var pdfTop = content.Height - (line.Box.Top * scaleY);
+ var pdfBottom = content.Height - (line.Box.Bottom * scaleY);
+ var size = Math.Max(4, pdfTop - pdfBottom);
+ var format = new HCharFormat(Size: (int)Math.Round(size * 100));
+ foreach (var w in line.Words)
+ {
+ var pdfLeft = w.Box.Left * scaleX;
+ var pdfRight = w.Box.Right * scaleX;
+ words.Add(new TextWord(CleanText(w.Text), pdfLeft, pdfRight, pdfTop, pdfBottom, pdfBottom, size, format, null));
+ }
+ }
+ }
+ }
+ catch
+ {
+ // Fall back to pure scan if OCR encounters an error
+ }
+
+ if (words.Count == 0)
+ {
+ content.Scanned = true;
+ if (!_textOnly)
+ content.ScanPath = scanPath;
+ return content;
+ }
+ }
+ }
+
+ if (words.Count == 0)
+ {
+ // A scan, a page of drawings, or a turned page: the rendered page keeps everything.
+ content.Scanned = true;
+ if (!_textOnly)
+ content.ScanPath = RenderScan(page.Number - 1);
+ return content;
+ }
}
// Plain text (TXT, page texts for slides) keeps table cells as paragraphs in reading order.
@@ -484,8 +569,9 @@ internal sealed partial class PdfDocumentReader
private string NewMediaPath(string extension)
{
- Directory.CreateDirectory(_mediaFolder);
- return Path.Combine(_mediaFolder, $"pdf-{++_pictures}.{extension}");
+ var folder = string.IsNullOrEmpty(_mediaFolder) ? Path.Combine(Path.GetTempPath(), "filee-media") : _mediaFolder;
+ Directory.CreateDirectory(folder);
+ return Path.Combine(folder, $"pdf-{++_pictures}.{extension}");
}
private SKBitmap Render(int pageIndex)
@@ -966,7 +1052,7 @@ internal sealed partial class PdfDocumentReader
private static string? NullIfBlank(string? text) => string.IsNullOrWhiteSpace(text) ? null : text.Trim();
- /// <summary>Plain text of a document read by <see cref="ReadText"/>: paragraphs separated by an empty line.</summary>
+ /// <summary>Plain text of a document read by <see cref="ReadText(string)"/>: paragraphs separated by an empty line.</summary>
public static string PlainText(HDocument document)
{
var sb = new StringBuilder();
diff --git a/src/Filee.Engines/Pdf/PdfTextConverter.cs b/src/Filee.Engines/Pdf/PdfTextConverter.cs
index 86859ffb..1588eed2 100644
--- a/src/Filee.Engines/Pdf/PdfTextConverter.cs
+++ b/src/Filee.Engines/Pdf/PdfTextConverter.cs
@@ -40,7 +40,7 @@ public sealed class PdfTextConverter : IConverter
private static IReadOnlyList<string> Text(ConversionStep step, IProgress<double>? progress, CancellationToken cancellationToken)
{
progress?.Report(0.1);
- var text = PdfDocumentReader.PlainText(PdfDocumentReader.ReadText(step.InputPath));
+ var text = PdfDocumentReader.PlainText(PdfDocumentReader.ReadText(step.InputPath, null, step.Preset.Document.OcrLanguage, step.Preset.Document.Ocr));
cancellationToken.ThrowIfCancellationRequested();
if (text.Length == 0)
throw NoText();
@@ -59,8 +59,8 @@ public sealed class PdfTextConverter : IConverter
progress?.Report(0.1);
var textOnly = step.Preset.Document.LeaveOutPictures;
var document = textOnly
- ? PdfDocumentReader.ReadText(step.InputPath)
- : PdfDocumentReader.Read(step.InputPath, Path.Combine(step.WorkDirectory, $"media-{Guid.NewGuid():N}"));
+ ? PdfDocumentReader.ReadText(step.InputPath, null, step.Preset.Document.OcrLanguage, step.Preset.Document.Ocr)
+ : PdfDocumentReader.Read(step.InputPath, Path.Combine(step.WorkDirectory, $"media-{Guid.NewGuid():N}"), null, step.Preset.Document.OcrLanguage, step.Preset.Document.Ocr);
cancellationToken.ThrowIfCancellationRequested();
if (textOnly && PdfDocumentReader.PlainText(document).Length == 0)
throw NoText();
diff --git a/src/Filee.Platform.MacOS/MacOsVisionOcrEngine.cs b/src/Filee.Platform.MacOS/MacOsVisionOcrEngine.cs
new file mode 100644
index 00000000..780f9e86
--- /dev/null
+++ b/src/Filee.Platform.MacOS/MacOsVisionOcrEngine.cs
@@ -0,0 +1,272 @@
+// OCR through Apple's Vision framework (VNRecognizeTextRequest), called directly via the Objective-C runtime so
+// Filee.Platform.MacOS stays free of an Objective-C bridging library. Zero install size and GPU/NPU accelerated;
+// the preferred engine on macOS whenever its SupportedLanguages cover what was asked for.
+
+using System.Runtime.InteropServices;
+using System.Runtime.Versioning;
+using System.Text;
+using Filee.Core.Ocr;
+
+namespace Filee.Platform.MacOS;
+
+/// <summary>Optical Character Recognition engine powered by Apple's Vision framework (macOS only).</summary>
+[SupportedOSPlatform("macos")]
+public sealed class MacOsVisionOcrEngine : IOcrEngine
+{
+ private static readonly Lazy<bool> VisionAvailable = new(() =>
+ {
+ if (!OperatingSystem.IsMacOS())
+ return false;
+ try
+ {
+ NativeLibrary.Load("/System/Library/Frameworks/Vision.framework/Vision");
+ return objc_getClass("VNRecognizeTextRequest") != 0;
+ }
+ catch
+ {
+ return false;
+ }
+ });
+
+ public string Id => "vision";
+
+ public string DisplayName => "Apple Vision";
+
+ public bool IsAvailable => VisionAvailable.Value;
+
+ public IReadOnlyList<string> SupportedLanguages { get; } =
+ [
+ "en-US", "fr-FR", "it-IT", "de-DE", "es-ES", "pt-BR", "zh-Hans", "zh-Hant", "ja-JP", "ko-KR", "ru-RU", "uk-UA"
+ ];
+
+ public Task<OcrPageResult> RecognizeAsync(string imagePath, string? language = null, CancellationToken cancellationToken = default)
+ {
+ if (!IsAvailable)
+ return Task.FromResult(OcrPageResult.Empty());
+
+ return Task.Run(() =>
+ {
+ cancellationToken.ThrowIfCancellationRequested();
+ using var pool = new AutoreleasePool();
+
+ var (width, height) = GetImageDimensions(imagePath);
+ if (width <= 0 || height <= 0)
+ (width, height) = (1000, 1000);
+
+ var nsPath = CFStringCreateWithCString(0, imagePath, 0x08000100);
+ if (nsPath == 0)
+ return OcrPageResult.Empty(width, height);
+
+ try
+ {
+ var urlClass = objc_getClass("NSURL");
+ var fileUrl = Send(urlClass, sel_registerName("fileURLWithPath:"), nsPath);
+
+ var reqClass = objc_getClass("VNRecognizeTextRequest");
+ var request = Send(Send(reqClass, sel_registerName("alloc")), sel_registerName("init"));
+ SendInteger(request, sel_registerName("setRecognitionLevel:"), 0); // 0 = VNRequestTextRecognitionLevelAccurate
+ SendBool(request, sel_registerName("setUsesLanguageCorrection:"), true);
+
+ if (!string.IsNullOrEmpty(language))
+ {
+ var langString = CFStringCreateWithCString(0, language, 0x08000100);
+ if (langString != 0)
+ {
+ var arrayClass = objc_getClass("NSArray");
+ var langsArray = Send(arrayClass, sel_registerName("arrayWithObject:"), langString);
+ Send(request, sel_registerName("setRecognitionLanguages:"), langsArray);
+ CFRelease(langString);
+ }
+ }
+
+ var dictClass = objc_getClass("NSDictionary");
+ var emptyDict = Send(dictClass, sel_registerName("dictionary"));
+ var handlerClass = objc_getClass("VNImageRequestHandler");
+ var handler = Send(Send(handlerClass, sel_registerName("alloc")), sel_registerName("initWithURL:options:"), fileUrl, emptyDict);
+
+ var arrayClass2 = objc_getClass("NSArray");
+ var requestsArray = Send(arrayClass2, sel_registerName("arrayWithObject:"), request);
+
+ if (!SendPerform(handler, sel_registerName("performRequests:error:"), requestsArray, out _))
+ return OcrPageResult.Empty(width, height);
+
+ var results = Send(request, sel_registerName("results"));
+ if (results == 0)
+ return OcrPageResult.Empty(width, height);
+
+ var count = (int)CFArrayGetCount(results);
+ var lines = new List<OcrLine>();
+ var allWords = new List<OcrWord>();
+ var fullText = new StringBuilder();
+
+ for (var i = 0; i < count; i++)
+ {
+ cancellationToken.ThrowIfCancellationRequested();
+ var obs = CFArrayGetValueAtIndex(results, i);
+ var candidates = Send(obs, sel_registerName("topCandidates:"), 1);
+ if (candidates == 0 || CFArrayGetCount(candidates) == 0)
+ continue;
+
+ var cand = CFArrayGetValueAtIndex(candidates, 0);
+ var nsStr = Send(cand, sel_registerName("string"));
+ var utf8Ptr = Send(nsStr, sel_registerName("UTF8String"));
+ var text = Marshal.PtrToStringUTF8(utf8Ptr);
+ if (string.IsNullOrWhiteSpace(text))
+ continue;
+
+ var conf = SendFloat(cand, sel_registerName("confidence"));
+ var normBox = SendRect(obs, sel_registerName("boundingBox"));
+
+ // Apple Vision returns normalized coordinates with bottom-left origin:
+ // normBox.X (0..1 from left), normBox.Y (0..1 from bottom),
+ // normBox.Width, normBox.Height.
+ // Convert to pixel coordinates with top-left origin:
+ var pixelLeft = normBox.X * width;
+ var pixelRight = (normBox.X + normBox.Width) * width;
+ var pixelTop = (1.0 - normBox.Y - normBox.Height) * height;
+ var pixelBottom = (1.0 - normBox.Y) * height;
+ var lineBox = new OcrRect(pixelLeft, pixelTop, pixelRight, pixelBottom);
+
+ // Estimate word boxes along the line
+ var rawWords = text.Split(' ', StringSplitOptions.RemoveEmptyEntries);
+ var wordsInLine = new List<OcrWord>();
+ if (rawWords.Length == 1)
+ {
+ var word = new OcrWord(rawWords[0], lineBox, conf);
+ wordsInLine.Add(word);
+ allWords.Add(word);
+ }
+ else if (rawWords.Length > 1)
+ {
+ var totalChars = rawWords.Sum(w => w.Length);
+ var currentX = pixelLeft;
+ var usableWidth = Math.Max(0, lineBox.Width);
+ for (var w = 0; w < rawWords.Length; w++)
+ {
+ var wStr = rawWords[w];
+ var wWidth = totalChars > 0 ? (usableWidth * wStr.Length / (double)totalChars) : (usableWidth / rawWords.Length);
+ var wBox = new OcrRect(currentX, pixelTop, currentX + wWidth, pixelBottom);
+ var word = new OcrWord(wStr, wBox, conf);
+ wordsInLine.Add(word);
+ allWords.Add(word);
+ currentX += wWidth;
+ }
+ }
+
+ lines.Add(new OcrLine(text, wordsInLine, lineBox));
+ fullText.AppendLine(text);
+ }
+
+ return new OcrPageResult(fullText.ToString().TrimEnd(), lines, allWords, width, height, language);
+ }
+ finally
+ {
+ CFRelease(nsPath);
+ }
+ }, cancellationToken);
+ }
+
+ private static (int Width, int Height) GetImageDimensions(string path)
+ {
+ try
+ {
+ var nsImageClass = objc_getClass("NSImage");
+ var nsPath = CFStringCreateWithCString(0, path, 0x08000100);
+ if (nsPath == 0)
+ return (0, 0);
+ try
+ {
+ var alloc = Send(nsImageClass, sel_registerName("alloc"));
+ var image = Send(alloc, sel_registerName("initWithContentsOfFile:"), nsPath);
+ if (image == 0)
+ return (0, 0);
+ var size = SendSize(image, sel_registerName("size"));
+ return ((int)Math.Round(size.Width), (int)Math.Round(size.Height));
+ }
+ finally
+ {
+ CFRelease(nsPath);
+ }
+ }
+ catch
+ {
+ return (0, 0);
+ }
+ }
+
+ [StructLayout(LayoutKind.Sequential)]
+ private struct CGRect
+ {
+ public double X;
+ public double Y;
+ public double Width;
+ public double Height;
+ }
+
+ private readonly struct AutoreleasePool : IDisposable
+ {
+ private readonly nint _pool = Send(Send(objc_getClass("NSAutoreleasePool"), sel_registerName("alloc")), sel_registerName("init"));
+ public AutoreleasePool() { }
+ public void Dispose() => SendVoid(_pool, sel_registerName("drain"));
+ }
+
+ [StructLayout(LayoutKind.Sequential)]
+ private struct CGSize
+ {
+ public double Width;
+ public double Height;
+ }
+
+ private const string CoreFoundation = "/System/Library/Frameworks/CoreFoundation.framework/CoreFoundation";
+ private const string ObjC = "/usr/lib/libobjc.A.dylib";
+
+ [DllImport(CoreFoundation)]
+ private static extern nint CFArrayGetCount(nint array);
+
+ [DllImport(CoreFoundation)]
+ private static extern nint CFArrayGetValueAtIndex(nint array, nint index);
+
+ [DllImport(CoreFoundation)]
+ private static extern nint CFStringCreateWithCString(nint allocator, [MarshalAs(UnmanagedType.LPUTF8Str)] string value, uint encoding);
+
+ [DllImport(CoreFoundation)]
+ private static extern void CFRelease(nint obj);
+
+ [DllImport(ObjC)]
+ private static extern nint objc_getClass([MarshalAs(UnmanagedType.LPUTF8Str)] string name);
+
+ [DllImport(ObjC)]
+ private static extern nint sel_registerName([MarshalAs(UnmanagedType.LPUTF8Str)] string name);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ private static extern nint Send(nint receiver, nint selector);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ private static extern nint Send(nint receiver, nint selector, nint argument);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ private static extern nint Send(nint receiver, nint selector, nint arg1, nint arg2);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ [return: MarshalAs(UnmanagedType.I1)]
+ private static extern bool SendBool(nint receiver, nint selector, [MarshalAs(UnmanagedType.I1)] bool argument);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ private static extern void SendInteger(nint receiver, nint selector, nint value);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ [return: MarshalAs(UnmanagedType.I1)]
+ private static extern bool SendPerform(nint receiver, nint selector, nint requests, out nint error);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ private static extern float SendFloat(nint receiver, nint selector);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ private static extern CGRect SendRect(nint receiver, nint selector);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ private static extern CGSize SendSize(nint receiver, nint selector);
+
+ [DllImport(ObjC, EntryPoint = "objc_msgSend")]
+ private static extern void SendVoid(nint receiver, nint selector);
+}
diff --git a/tests/Filee.App.Tests/EnginesPageTests.cs b/tests/Filee.App.Tests/EnginesPageTests.cs
index f1c6edec..dcc491a4 100644
--- a/tests/Filee.App.Tests/EnginesPageTests.cs
+++ b/tests/Filee.App.Tests/EnginesPageTests.cs
@@ -111,4 +111,17 @@ public class EnginesPageTests
vm.Dispose();
}
}
+
+ [AvaloniaFact]
+ public void Tesseract_package_has_install_available_when_not_installed()
+ {
+ TestServices.EnsureInitialized("en");
+ var downloads = AppHost.Get<EngineDownloadService>();
+ var tesseract = downloads.Packages.FirstOrDefault(p => p.Package.Id == "tesseract");
+ Assert.NotNull(tesseract);
+ if (EnginePackageState.CanSystemInstall(tesseract.Package))
+ {
+ Assert.True(tesseract.CanInstall || tesseract.IsInstalled);
+ }
+ }
}
diff --git a/tests/Filee.Core.Tests/OcrEngineRegistryTests.cs b/tests/Filee.Core.Tests/OcrEngineRegistryTests.cs
new file mode 100644
index 00000000..1a6f16e6
--- /dev/null
+++ b/tests/Filee.Core.Tests/OcrEngineRegistryTests.cs
@@ -0,0 +1,88 @@
+using Filee.Core.Ocr;
+using Filee.Core.Presets;
+
+namespace Filee.Core.Tests;
+
+public class OcrEngineRegistryTests
+{
+ private sealed class MockOcrEngine(string id, string name, bool available, IReadOnlyList<string> languages) : IOcrEngine
+ {
+ public string Id => id;
+ public string DisplayName => name;
+ public bool IsAvailable => available;
+ public IReadOnlyList<string> SupportedLanguages => languages;
+
+ public Task<OcrPageResult> RecognizeAsync(string imagePath, string? language = null, CancellationToken cancellationToken = default)
+ {
+ var word = new OcrWord("test", new OcrRect(0, 0, 10, 10), 0.99f);
+ var line = new OcrLine("test", [word], new OcrRect(0, 0, 10, 10));
+ return Task.FromResult(new OcrPageResult("test", [line], [word], 100, 100, language));
+ }
+ }
+
+ [Fact]
+ public void OcrRect_computes_width_and_height_correctly()
+ {
+ var rect = new OcrRect(10, 20, 110, 70);
+ Assert.Equal(100, rect.Width);
+ Assert.Equal(50, rect.Height);
+ }
+
+ [Fact]
+ public void OcrPageResult_Empty_has_zero_words_and_lines()
+ {
+ var empty = OcrPageResult.Empty(200, 300);
+ Assert.Empty(empty.Text);
+ Assert.Empty(empty.Lines);
+ Assert.Empty(empty.Words);
+ Assert.Equal(200, empty.ImageWidth);
+ Assert.Equal(300, empty.ImageHeight);
+ }
+
+ [Fact]
+ public void Register_and_retrieve_engine()
+ {
+ var mock = new MockOcrEngine("mock-ocr", "Mock OCR", true, ["en", "ne"]);
+ OcrEngineRegistry.Register(mock);
+
+ Assert.Contains(OcrEngineRegistry.All, e => e.Id == "mock-ocr");
+ Assert.NotNull(OcrEngineRegistry.GetById("mock-ocr"));
+ Assert.Same(mock, OcrEngineRegistry.GetById("mock-ocr"));
+ }
+
+ [Fact]
+ public void DocumentOptions_defaults_and_cloning()
+ {
+ var options = new DocumentOptions();
+ Assert.True(options.Ocr);
+ Assert.Null(options.OcrLanguage);
+
+ options.Ocr = false;
+ options.OcrLanguage = "ne";
+
+ var clone = options.Clone();
+ Assert.False(clone.Ocr);
+ Assert.Equal("ne", clone.OcrLanguage);
+ }
+
+ [Fact]
+ public void GetDefault_routes_a_language_to_the_engine_that_actually_supports_it()
+ {
+ // A language no other test's mock claims, so this isn't order-dependent on the shared registry.
+ var tesseract = new MockOcrEngine("tesseract", "Tesseract OCR", true, ["th", "el"]);
+ OcrEngineRegistry.Register(tesseract);
+
+ var thaiEngine = OcrEngineRegistry.GetDefault("th");
+ Assert.NotNull(thaiEngine);
+ Assert.Equal("tesseract", thaiEngine.Id);
+ }
+
+ [Fact]
+ public void GetDefault_returns_null_when_no_available_engine_supports_the_language()
+ {
+ var unrelated = new MockOcrEngine("unrelated-mock", "Unrelated", true, ["fr"]);
+ OcrEngineRegistry.Register(unrelated);
+
+ Assert.Null(OcrEngineRegistry.GetDefault("a-language-nothing-claims"));
+ }
+}
diff --git a/tests/Filee.Engines.Tests/TesseractOcrTests.cs b/tests/Filee.Engines.Tests/TesseractOcrTests.cs
new file mode 100644
index 00000000..b5fe537a
--- /dev/null
+++ b/tests/Filee.Engines.Tests/TesseractOcrTests.cs
@@ -0,0 +1,74 @@
+using Filee.Core.Ocr;
+using Filee.Engines.Ocr;
+using SkiaSharp;
+
+namespace Filee.Engines.Tests;
+
+public sealed class TesseractOcrTests
+{
+ // Pure mapping logic: runs even where Tesseract itself isn't installed (CI).
+ [Theory]
+ [InlineData("en", "eng")]
+ [InlineData("ne", "nep")]
+ [InlineData("zh-Hans", "chi_sim")]
+ [InlineData("zh-Hant", "chi_tra")]
+ [InlineData(null, "eng")]
+ public void MapLanguage_maps_known_codes(string? filee, string tesseract) =>
+ Assert.Equal(tesseract, TesseractOcrEngine.MapLanguage(filee));
+
+ [Fact]
+ public void MapLanguage_returns_null_for_a_code_with_no_trained_data()
+ {
+ // Unlike the old PaddleOCR engine, which silently fell back to English for any code it didn't
+ // recognize, an unmapped code must be reported as unsupported rather than guessed at.
+ Assert.Null(TesseractOcrEngine.MapLanguage("xal"));
+ }
+
+ [Fact]
+ public async Task RecognizesTextInAGeneratedImage()
+ {
+ var engine = new TesseractOcrEngine();
+ Assert.SkipUnless(engine.IsAvailable, "Tesseract is not installed on this machine.");
+
+ var imagePath = Path.Combine(Path.GetTempPath(), $"filee-tesseract-test-{Guid.NewGuid():N}.png");
+ try
+ {
+ using (var bitmap = new SKBitmap(400, 120))
+ using (var canvas = new SKCanvas(bitmap))
+ using (var font = new SKFont(SKTypeface.Default, 48))
+ using (var paint = new SKPaint { Color = SKColors.Black, IsAntialias = true })
+ {
+ canvas.Clear(SKColors.White);
+ canvas.DrawText("Filee OCR Test", 10, 70, font, paint);
+ using var image = SKImage.FromBitmap(bitmap);
+ using var data = image.Encode(SKEncodedImageFormat.Png, 100);
+ using var stream = File.OpenWrite(imagePath);
+ data.SaveTo(stream);
+ }
+
+ var result = await engine.RecognizeAsync(imagePath, "en", TestContext.Current.CancellationToken);
+
+ Assert.Contains("Filee", result.Text, StringComparison.OrdinalIgnoreCase);
+ Assert.NotEmpty(result.Words);
+ Assert.True(result.ImageWidth > 0 && result.ImageHeight > 0);
+ }
+ finally
+ {
+ File.Delete(imagePath);
+ }
+ }
+
+ [Fact]
+ public async Task RecognizeAsync_throws_a_clear_error_for_an_unmapped_language()
+ {
+ var engine = new TesseractOcrEngine();
+ Assert.SkipUnless(engine.IsAvailable, "Tesseract is not installed on this machine.");
+
+ // Any existing file: the unmapped-language check happens before the image is read.
+ var imagePath = Path.Combine(AppContext.BaseDirectory, "Filee.Engines.Tests.dll");
+
+ var ex = await Assert.ThrowsAsync<InvalidOperationException>(() =>
+ engine.RecognizeAsync(imagePath, "xal", TestContext.Current.CancellationToken));
+ Assert.Contains("xal", ex.Message);
+ }
+}
diff --git a/tests/Filee.Platform.Tests/Filee.Platform.Tests.csproj b/tests/Filee.Platform.Tests/Filee.Platform.Tests.csproj
index a3d0e81a..27f0ce7f 100644
--- a/tests/Filee.Platform.Tests/Filee.Platform.Tests.csproj
+++ b/tests/Filee.Platform.Tests/Filee.Platform.Tests.csproj
@@ -8,6 +8,7 @@
<ProjectReference Include="..\..\src\Filee.Platform.Linux\Filee.Platform.Linux.csproj" />
<ProjectReference Include="..\..\src\Filee.Platform.MacOS\Filee.Platform.MacOS.csproj" />
<ProjectReference Include="..\..\src\Filee.Platform.Windows\Filee.Platform.Windows.csproj" />
+ <ProjectReference Include="..\..\src\Filee.Engines\Filee.Engines.csproj" />
<PackageReference Include="Microsoft.NET.Test.Sdk" />
<PackageReference Include="xunit.v3" />
<PackageReference Include="xunit.runner.visualstudio" />
diff --git a/tests/Filee.Platform.Tests/VisionOcrTests.cs b/tests/Filee.Platform.Tests/VisionOcrTests.cs
new file mode 100644
index 00000000..de906b16
--- /dev/null
+++ b/tests/Filee.Platform.Tests/VisionOcrTests.cs
@@ -0,0 +1,198 @@
+using Filee.Core.Conversion;
+using Filee.Core.Ocr;
+using Filee.Core.Presets;
+using Filee.Engines.Hwp.Hwpx;
+using Filee.Engines.Ocr;
+using Filee.Engines.Pdf;
+using Filee.Platform.MacOS;
+
+namespace Filee.Platform.Tests;
+
+public sealed class VisionOcrTests
+{
+ private static string FindRepoRoot()
+ {
+ var dir = new DirectoryInfo(AppContext.BaseDirectory);
+ while (dir is not null && !File.Exists(Path.Combine(dir.FullName, "Filee.slnx")))
+ dir = dir.Parent;
+ return dir!.FullName;
+ }
+
+ [Fact]
+ public async Task VisionOcrRecognizesTextInImageOnMacOS()
+ {
+ if (!OperatingSystem.IsMacOS())
+ {
+ Assert.Skip("Requires macOS Vision framework.");
+ return;
+ }
+
+ var engine = new MacOsVisionOcrEngine();
+ Assert.True(engine.IsAvailable);
+ Assert.Equal("vision", engine.Id);
+ Assert.NotEmpty(engine.SupportedLanguages);
+
+ var imagePath = Path.Combine(FindRepoRoot(), "docs", "media", "banner-en.jpg");
+ Assert.True(File.Exists(imagePath), $"Test image {imagePath} should exist");
+
+ var result = await engine.RecognizeAsync(imagePath, "en-US", TestContext.Current.CancellationToken);
+
+ Assert.NotEmpty(result.Text);
+ Assert.NotEmpty(result.Lines);
+ Assert.NotEmpty(result.Words);
+ Assert.True(result.ImageWidth > 0);
+ Assert.True(result.ImageHeight > 0);
+
+ // The banner image has "Filee"
+ Assert.Contains("Filee", result.Text, StringComparison.OrdinalIgnoreCase);
+ }
+
+ [Fact]
+ public async Task ImageOcrConverterConvertsImageToTextAndDocx()
+ {
+ if (!OperatingSystem.IsMacOS())
+ {
+ Assert.Skip("Requires macOS Vision framework.");
+ return;
+ }
+
+ var engine = new MacOsVisionOcrEngine();
+ OcrEngineRegistry.Register(engine);
+
+ var converter = new ImageOcrConverter();
+ Assert.True(converter.GetStatus().IsAvailable);
+
+ var imagePath = Path.Combine(FindRepoRoot(), "docs", "media", "banner-en.jpg");
+ var workDir = Path.Combine(Path.GetTempPath(), $"filee-test-{Guid.NewGuid():N}");
+ Directory.CreateDirectory(workDir);
+
+ try
+ {
+ var txtOut = Path.Combine(workDir, "banner.txt");
+ var txtStep = new ConversionStep(
+ InputPath: imagePath,
+ From: "jpg",
+ To: "txt",
+ Preset: new Preset(),
+ Output: new FixedPath(txtOut),
+ WorkDirectory: workDir);
+
+ var txtOutputs = await converter.ConvertAsync(txtStep, null, TestContext.Current.CancellationToken);
+ Assert.Single(txtOutputs);
+ Assert.True(File.Exists(txtOutputs[0]));
+ var txt = await File.ReadAllTextAsync(txtOutputs[0], TestContext.Current.CancellationToken);
+ Assert.Contains("Filee", txt, StringComparison.OrdinalIgnoreCase);
+
+ var docxOut = Path.Combine(workDir, "banner.docx");
+ var docxStep = new ConversionStep(
+ InputPath: imagePath,
+ From: "jpg",
+ To: "docx",
+ Preset: new Preset(),
+ Output: new FixedPath(docxOut),
+ WorkDirectory: workDir);
+
+ var docxOutputs = await converter.ConvertAsync(docxStep, null, TestContext.Current.CancellationToken);
+ Assert.Single(docxOutputs);
+ Assert.True(File.Exists(docxOutputs[0]));
+ Assert.True(new FileInfo(docxOutputs[0]).Length > 1000);
+ }
+ finally
+ {
+ Directory.Delete(workDir, recursive: true);
+ }
+ }
+
+ [Fact]
+ public void PdfDocumentReaderExtractsEditableTextFromScannedPdf()
+ {
+ if (!OperatingSystem.IsMacOS())
+ {
+ Assert.Skip("Requires macOS Vision framework.");
+ return;
+ }
+
+ var engine = new MacOsVisionOcrEngine();
+ OcrEngineRegistry.Register(engine);
+
+ var imagePath = Path.Combine(FindRepoRoot(), "docs", "media", "banner-en.jpg");
+ var workDir = Path.Combine(Path.GetTempPath(), $"filee-pdf-test-{Guid.NewGuid():N}");
+ Directory.CreateDirectory(workDir);
+
+ try
+ {
+ var pdfPath = Path.Combine(workDir, "scanned.pdf");
+
+ // Create a pure scanned PDF with only a bitmap image using PdfSharp
+ using (var pdfDoc = new PdfSharp.Pdf.PdfDocument())
+ {
+ var page = pdfDoc.AddPage();
+ page.Width = PdfSharp.Drawing.XUnit.FromPoint(600);
+ page.Height = PdfSharp.Drawing.XUnit.FromPoint(400);
+ using var gfx = PdfSharp.Drawing.XGraphics.FromPdfPage(page);
+ using var img = PdfSharp.Drawing.XImage.FromFile(imagePath);
+ gfx.DrawImage(img, 0, 0, 600, 400);
+ pdfDoc.Save(pdfPath);
+ }
+
+ Assert.True(File.Exists(pdfPath));
+
+ // Read the scanned PDF using PdfDocumentReader with OCR enabled
+ var mediaFolder = Path.Combine(workDir, "media");
+ var document = PdfDocumentReader.Read(pdfPath, mediaFolder, engine);
+
+ Assert.NotEmpty(document.Sections);
+ var paragraphs = document.Sections.SelectMany(s => s.Blocks).OfType<HParagraph>().ToList();
+ Assert.NotEmpty(paragraphs);
+
+ var allText = string.Join(" ", paragraphs.SelectMany(p => p.Inlines).OfType<HText>().Select(t => t.Text));
+ Assert.Contains("Filee", allText, StringComparison.OrdinalIgnoreCase);
+
+ // Verify plain text reader also extracts the OCR text
+ var plainDoc = PdfDocumentReader.ReadText(pdfPath, engine);
+ var plainText = PdfDocumentReader.PlainText(plainDoc);
+ Assert.Contains("Filee", plainText, StringComparison.OrdinalIgnoreCase);
+ }
+ finally
+ {
+ Directory.Delete(workDir, recursive: true);
+ }
+ }
+
+ [Fact]
+ public void ScannedPdfWithoutOcrStaysPictureOnly()
+ {
+ var imagePath = Path.Combine(FindRepoRoot(), "docs", "media", "banner-en.jpg");
+ var workDir = Path.Combine(Path.GetTempPath(), $"filee-pdf-noocr-{Guid.NewGuid():N}");
+ Directory.CreateDirectory(workDir);
+
+ try
+ {
+ var pdfPath = Path.Combine(workDir, "scanned.pdf");
+ using (var pdfDoc = new PdfSharp.Pdf.PdfDocument())
+ {
+ var page = pdfDoc.AddPage();
+ page.Width = PdfSharp.Drawing.XUnit.FromPoint(600);
+ page.Height = PdfSharp.Drawing.XUnit.FromPoint(400);
+ using var gfx = PdfSharp.Drawing.XGraphics.FromPdfPage(page);
+ using var img = PdfSharp.Drawing.XImage.FromFile(imagePath);
+ gfx.DrawImage(img, 0, 0, 600, 400);
+ pdfDoc.Save(pdfPath);
+ }
+
+ var document = PdfDocumentReader.Read(pdfPath, Path.Combine(workDir, "media"), null, null, enableOcr: false);
+ var paragraphs = document.Sections.SelectMany(s => s.Blocks).OfType<HParagraph>()
+ .SelectMany(p => p.Inlines).OfType<HText>().Select(t => t.Text).ToList();
+ Assert.Empty(paragraphs);
+ }
+ finally
+ {
+ Directory.Delete(workDir, recursive: true);
+ }
+ }
+
+ private sealed class FixedPath(string path) : IOutputAllocator
+ {
+ public string? Allocate(string extension, string? suffix = null) => path;
+ }
+}