Voice-to-text dictation app with local (Nvidia Parakeet/Whisper) and cloud models (BYOK). Privacy-first and available cross-platform.
JavaScript aianthropiccross-platformgeminigroqlinuxmacosnvidiaopen-sourceopenaiparakeetspeech-to-texttranscribewhisperwindows
fix: newline issues
Gabriel Steinea61d246e51a
diff --git a/src/helpers/whisper.js b/src/helpers/whisper.jsindex de4b5b8e4..f9ed8d6cb 100644--- a/src/helpers/whisper.js+++ b/src/helpers/whisper.js@@ -758,6 +758,12 @@ class WhisperManager {}}+ // Normalize whitespace: replace newlines with spaces and collapse multiple spaces+ // whisper.cpp returns text with \n between audio segments which causes formatting issues+ normalizeWhitespace(text) {+ return text.replace(/\n/g, " ").replace(/\s+/g, " ").trim();+ }+parseWhisperResult(output) {// Handle both string (from CLI) and object (from server) inputslet result;@@ -767,7 +773,7 @@ class WhisperManager {result = JSON.parse(output);} catch (parseError) {// Try parsing as plain text (non-JSON output)- const text = output.trim();+ const text = this.normalizeWhitespace(output);if (text && !this.isBlankAudioMarker(text)) {return { success: true, text };}@@ -782,10 +788,9 @@ class WhisperManager {// Handle whisper.cpp JSON format (CLI mode)if (result.transcription && Array.isArray(result.transcription)) {- const text = result.transcription- .map((seg) => seg.text)- .join("")- .trim();+ const text = this.normalizeWhitespace(+ result.transcription.map((seg) => seg.text).join("")+ );if (!text || this.isBlankAudioMarker(text)) {return { success: false, message: "No audio detected" };}@@ -794,7 +799,8 @@ class WhisperManager {// Handle whisper-server format (has "text" field directly)if (result.text !== undefined) {- const text = typeof result.text === "string" ? result.text.trim() : "";+ const text =+ typeof result.text === "string" ? this.normalizeWhitespace(result.text) : "";if (!text || this.isBlankAudioMarker(text)) {return { success: false, message: "No audio detected" };}