Skip to content

The truly free, open-source, local-first, cross-platform CapCut replacement. Scriptable through a JSON-RPC API & MCP.

feat(captions, speech): a transcription's and a read's bar moves through each stage and says how long is left

Jareerc71f9001ed7c
diff --git a/src/crates/concat-speech/src/lib.rs b/src/crates/concat-speech/src/lib.rs
index 83cd34b3..d7cc5103 100644
--- a/src/crates/concat-speech/src/lib.rs
+++ b/src/crates/concat-speech/src/lib.rs
@@ -40,6 +40,29 @@ pub const fn accelerated() -> bool {
cfg!(any(target_os = "macos", target_os = "windows"))
}
+/// How far a transcription or a read has got, as it reports it: the stage
+/// it is in, and in the work itself where it is and where the step under
+/// way will leave it. A job is mostly long waits between reports - whisper
+/// reports once per thirty seconds of sound, sherpa once per sentence - and
+/// `next` is what lets a bar move through the wait without promising more
+/// than the job has done.
+#[derive(Clone, Copy, Debug, PartialEq)]
+pub enum JobStage {
+ /// Reading the sound in: a clip's, or the recording a voice is read
+ /// from.
+ Reading,
+ /// Loading the model. Not reported when it is still loaded from the
+ /// run before.
+ Loading,
+ /// The work itself.
+ Working {
+ /// How much of it is finished, `0..=1`.
+ done: f32,
+ /// Where the step under way ends, `done..=1`.
+ next: f32,
+ },
+}
+
/// Progress for one model download.
#[derive(Clone, Debug, PartialEq, Eq)]
pub struct DownloadProgress {
diff --git a/src/crates/concat-speech/src/transcribe.rs b/src/crates/concat-speech/src/transcribe.rs
index 0a400c78..50f3276c 100644
--- a/src/crates/concat-speech/src/transcribe.rs
+++ b/src/crates/concat-speech/src/transcribe.rs
@@ -28,7 +28,7 @@ use concat_media::{AudioDecoder, AudioOptions, SampleFormat};
use serde::{Deserialize, Serialize};
use whisper_rs::{FullParams, SamplingStrategy, WhisperContext, WhisperContextParameters};
-use crate::DownloadProgress;
+use crate::{DownloadProgress, JobStage};
/// One downloadable Whisper model. Sizes are approximate, for progress bars
/// when the server declines to send a Content-Length.
@@ -307,9 +307,17 @@ impl Transcriber {
.map_err(|error| format!("could not remove {}: {error}", file.display()))
}
+ /// The model still loaded from the last run, which the next run on it
+ /// does not load again. None while a run holds it.
+ pub fn loaded_model(&self) -> Option<String> {
+ let loaded = self.model.try_lock().ok()?;
+ loaded.as_ref().map(|model| model.id.clone())
+ }
+
/// Transcribes one clip's audio window into caption segments. Blocks
- /// for the whole run: run it on its own thread. `progress` is called
- /// with a 0..100 percentage as whisper works.
+ /// for the whole run: run it on its own thread. `progress` is told each
+ /// stage as it begins, and in whisper's work each window it starts:
+ /// whisper reads thirty seconds at a time, so `next` is a window on.
///
/// Timestamps come back relative to the window: 0 is `source_start`. The
/// caller owns the clip, so the map onto the timeline (divide by speed,
@@ -318,7 +326,7 @@ impl Transcriber {
&self,
dirs: &AppDirs,
request: &TranscribeRequest,
- progress: impl FnMut(i32) + Send + 'static,
+ mut progress: impl FnMut(JobStage) + Send + 'static,
) -> Result<Vec<Segment>, String> {
let job = self.gate.begin("transcription")?;
let cancel = job.cancel_handle();
@@ -336,6 +344,7 @@ impl Transcriber {
}
// The window, as whisper wants it: 16 kHz mono floats.
+ progress(JobStage::Reading);
let mut decoder = AudioDecoder::open(
&request.path,
&AudioOptions {
@@ -363,6 +372,7 @@ impl Transcriber {
.map_err(|_| "transcriber state poisoned".to_owned())?;
if loaded.as_ref().map(|model| model.id.as_str()) != Some(request.model_id.as_str()) {
// Load before overwriting: a failed load keeps the old model.
+ progress(JobStage::Loading);
let context =
WhisperContext::new_with_params(&model_path, WhisperContextParameters::default())
.map_err(|error| format!("could not load the model: {error}"))?;
@@ -410,7 +420,20 @@ impl Transcriber {
params.set_abort_callback(Some(abort_when_set));
params.set_abort_callback_user_data(Arc::as_ptr(&cancel) as *mut std::ffi::c_void);
}
- params.set_progress_callback_safe(progress);
+ // Whisper calls back as it starts each window, with how far into
+ // the sound it is; the window it starts reads thirty seconds more.
+ let window = (30.0 / request.window).min(1.0) as f32;
+ progress(JobStage::Working {
+ done: 0.0,
+ next: window,
+ });
+ params.set_progress_callback_safe(move |percent: i32| {
+ let done = (percent as f32 / 100.0).clamp(0.0, 1.0);
+ progress(JobStage::Working {
+ done,
+ next: (done + window).min(1.0),
+ });
+ });
state
.full(params, &samples)
@@ -521,7 +544,7 @@ mod real_audio {
model_id: "tiny.en".to_owned(),
};
let segments = transcriber
- .transcribe(&dirs, &request, |percent| eprintln!("{percent}%"))
+ .transcribe(&dirs, &request, |stage| eprintln!("{stage:?}"))
.expect("transcription");
for segment in &segments {
eprintln!("{:.2}-{:.2} {}", segment.start, segment.end, segment.text);
diff --git a/src/crates/concat-speech/src/tts.rs b/src/crates/concat-speech/src/tts.rs
index e64c90c6..3c4a4cca 100644
--- a/src/crates/concat-speech/src/tts.rs
+++ b/src/crates/concat-speech/src/tts.rs
@@ -48,7 +48,7 @@ use sherpa_onnx::{
OfflineTtsModelConfig, OfflineTtsPocketModelConfig,
};
-use crate::DownloadProgress;
+use crate::{DownloadProgress, JobStage};
/// Which network a bundle is, and so which voices it reads with and how
/// the engine is configured for it.
@@ -704,14 +704,23 @@ impl Speech {
.map_err(|error| format!("could not remove {}: {error}", dir.display()))
}
+ /// The model still loaded from the last read, which the next read
+ /// with it does not load again. None while a read holds it.
+ pub fn loaded_model(&self) -> Option<String> {
+ let engine = self.engine.try_lock().ok()?;
+ engine.as_ref().map(|cached| cached.model_id.clone())
+ }
+
/// Synthesizes one narration clip into `<project>/audio/`. Blocks for
- /// the whole synthesis: run it on its own thread. `progress` is called
- /// with a 0..1 fraction as sentences complete.
+ /// the whole synthesis: run it on its own thread. `progress` is told
+ /// each stage as it begins, and in the reading itself how far it is:
+ /// Chatterbox's as it writes, Kokoro's and Pocket's as each sentence
+ /// is done.
pub fn speak(
&self,
dirs: &AppDirs,
request: &SpeakRequest,
- progress: impl FnMut(f32) + Send + 'static,
+ mut progress: impl FnMut(JobStage) + Send + 'static,
) -> Result<SpeakResult, String> {
let job = self.gate.begin("speech generation")?;
let cancel = job.cancel_handle();
@@ -792,6 +801,7 @@ impl Speech {
.is_none_or(|cached| cached.model_id != request.model_id);
if stale {
// Load before overwriting: a failed load keeps the old engine.
+ progress(JobStage::Loading);
let loading = std::time::Instant::now();
let load = |accelerated: bool| -> Result<Engine, String> {
log::info!(
@@ -873,15 +883,22 @@ impl Speech {
#[cfg(feature = "chatterbox")]
let tts = match tts {
Engine::Chatterbox(chatterbox) => {
+ progress(JobStage::Reading);
let samples = reference_samples(&recordings, request)?;
- let mut progress = progress;
+ progress(JobStage::Working {
+ done: 0.0,
+ next: CHATTERBOX_STEP,
+ });
let (spoken, on_cpu) = read_or_on_cpu(
chatterbox.as_ref(),
accelerated,
&request.model_id,
|chatterbox| {
- chatterbox.speak(text, &samples, &cancel, &mut |fraction| {
- progress(fraction);
+ chatterbox.speak(text, &samples, &cancel, &mut |done| {
+ progress(JobStage::Working {
+ done,
+ next: (done + CHATTERBOX_STEP).min(1.0),
+ });
})
},
|| crate::chatterbox::Engine::load(&dir, false),
@@ -931,6 +948,7 @@ impl Speech {
..Default::default()
},
_ => {
+ progress(JobStage::Reading);
let samples = reference_samples(&recordings, request)?;
GenerationConfig {
speed,
@@ -951,9 +969,17 @@ impl Speech {
}
};
+ // Sherpa reports as each sentence is done, and the next is about
+ // as long as the average one so far: before the first report, one
+ // of the sentences the text appears to have.
+ let first = 1.0 / sentences(text) as f32;
+ progress(JobStage::Working {
+ done: 0.0,
+ next: first,
+ });
let progress_cancel = Arc::clone(&cancel);
- let mut progress = progress;
let mut last_fraction = -1.0f32;
+ let mut reports = 0u32;
let audio = tts
.generate_with_config(
text,
@@ -964,9 +990,14 @@ impl Speech {
}
// The engine reports once per sentence; only meaningful
// movement is worth a redraw.
+ reports += 1;
if fraction - last_fraction >= 0.01 {
last_fraction = fraction;
- progress(fraction);
+ let step = fraction / reports as f32;
+ progress(JobStage::Working {
+ done: fraction,
+ next: (fraction + step).min(1.0),
+ });
}
true
}),
@@ -1028,6 +1059,21 @@ fn recordings_dir(dirs: &AppDirs) -> Result<PathBuf, String> {
model_dir(dirs, POCKET_ID)
}
+/// How far ahead of Chatterbox's last report its next is taken to be: it
+/// reports every few tokens, so a short way.
+const CHATTERBOX_STEP: f32 = 0.02;
+
+/// How many sentences `text` appears to have, at least one: the stretches
+/// with words in them between full stops, question and exclamation marks
+/// (and their CJK forms) and line breaks. An estimate, for a bar, of how
+/// sherpa will split it.
+fn sentences(text: &str) -> usize {
+ text.split(['.', '!', '?', '。', '!', '?', '\n'])
+ .filter(|sentence| sentence.chars().any(char::is_alphanumeric))
+ .count()
+ .max(1)
+}
+
/// The recording a request reads the voice from, as mono samples at
/// [`REFERENCE_RATE`]: one of the stock recordings in `dir` for a named
/// voice, or the request's reference for a chosen one, cut to
@@ -1214,6 +1260,18 @@ fn load_kokoro(dir: &Path, accelerated: bool) -> Result<OfflineTts, String> {
mod tests {
use super::*;
+ /// The sentences a bar expects are the stretches with words in them:
+ /// trailing marks, ellipses and blank lines add none, and a text with
+ /// no mark at all is one.
+ #[test]
+ fn sentences_are_the_stretches_with_words() {
+ assert_eq!(sentences("Hello there. How are you? Fine!"), 3);
+ assert_eq!(sentences("Wait... what?!\n\nOkay."), 3);
+ assert_eq!(sentences("no marks at all"), 1);
+ assert_eq!(sentences("。!"), 1);
+ assert_eq!(sentences("你好。再见!"), 2);
+ }
+
/// A read the accelerator's runtime fails is read again once on the
/// CPU, and the CPU engine comes back to be kept: DirectML refused
/// Chatterbox's attention mid-read, after it had loaded (#291).
diff --git a/src/crates/concat/locales/ar.json b/src/crates/concat/locales/ar.json
index a1d72cd6..d34ec87e 100644
--- a/src/crates/concat/locales/ar.json
+++ b/src/crates/concat/locales/ar.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "عنوان لكل عبارة منطوقة، مقصوص على توقيت الكلمات نفسها، على أول مسار فيه متّسع.",
"captions.oneTitlePerSentence": "عنوان لكل جملة، يبقى طوال المدة اللازمة لقراءته، واحدًا تلو الآخر على أول مسار فيه متّسع.",
"captions.placement": "الموضع",
+ "captions.readingTheSound": "جارٍ قراءة صوت المقطع",
"captions.transcribing": "جارٍ التفريغ النصي",
"captions.typeWordsPutScreen": "اكتب الكلمات التي ستظهر على الشاشة…",
"captions.useVoiceFromTimeline": "استخدام الصوت من المخطط الزمني",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "إلى {0}",
"clipInspector.treatments": "المعالجات",
"clipInspector.volume": "مستوى الصوت",
+ "common.aboutLeft": "بقي نحو {0}:{1}",
"common.adjust": "ضبط",
"common.apply": "تطبيق",
"common.audio": "الصوت",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "استورد أولًا مقطعًا فيه صوت.",
"speech.italian": "إيطالي",
"speech.japanese": "ياباني",
+ "speech.loadingTheModel": "جارٍ تحميل نموذج الصوت",
"speech.long": "طويلة",
"speech.longer": "أطول",
"speech.male": "ذكر",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "اختر عيّنة فيها صوت",
"speech.portuguese": "برتغالي",
"speech.quick": "سريع",
+ "speech.readingTheSample": "جارٍ قراءة تسجيل الصوت",
"speech.readsAnyVoiceFew": "يقرأ بأي صوت من بضع ثوانٍ من تسجيل: واحد مرفق به، أو عيّنة من الحاوية.",
"speech.readsVoiceHeardClip": "يقرأ بالصوت المسموع في المقطع المحدّد على المخطط الزمني",
"speech.readsVoiceHeardRecording": "يقرأ بالصوت المسموع في تسجيل من الحاوية. الثواني العشر الأولى من النقطة المختارة أدناه هي ما يستمع إليه النموذج.",
diff --git a/src/crates/concat/locales/bn.json b/src/crates/concat/locales/bn.json
index 970cffa7..318f2791 100644
--- a/src/crates/concat/locales/bn.json
+++ b/src/crates/concat/locales/bn.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "প্রতিটি বলা বাক্যাংশের জন্য একটি করে টাইটেল, শব্দের নিজস্ব সময় অনুযায়ী কাটা, জায়গা আছে এমন প্রথম লেনে।",
"captions.oneTitlePerSentence": "প্রতিটি বাক্যের জন্য একটি করে টাইটেল, পড়তে যতক্ষণ লাগে ততক্ষণ ধরে রাখা হয়, জায়গা আছে এমন প্রথম লেনে একের পর এক।",
"captions.placement": "অবস্থান",
+ "captions.readingTheSound": "ক্লিপের শব্দ পড়া হচ্ছে",
"captions.transcribing": "ট্রান্সক্রাইব করা হচ্ছে",
"captions.typeWordsPutScreen": "স্ক্রিনে দেখানোর শব্দগুলো লিখুন…",
"captions.useVoiceFromTimeline": "টাইমলাইনের কণ্ঠ ব্যবহার করুন",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "{0}-এ প্রবেশ",
"clipInspector.treatments": "ট্রিটমেন্ট",
"clipInspector.volume": "ভলিউম",
+ "common.aboutLeft": "আর প্রায় {0}:{1} বাকি",
"common.adjust": "সমন্বয় করুন",
"common.apply": "প্রয়োগ করুন",
"common.audio": "অডিও",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "আগে কণ্ঠ আছে এমন একটি ক্লিপ ইমপোর্ট করুন।",
"speech.italian": "ইতালীয়",
"speech.japanese": "জাপানি",
+ "speech.loadingTheModel": "কণ্ঠ মডেল লোড হচ্ছে",
"speech.long": "লম্বা",
"speech.longer": "আরও লম্বা",
"speech.male": "পুরুষ",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "কণ্ঠ আছে এমন একটি নমুনা বেছে নিন",
"speech.portuguese": "পর্তুগিজ",
"speech.quick": "দ্রুত",
+ "speech.readingTheSample": "কণ্ঠের রেকর্ডিং পড়া হচ্ছে",
"speech.readsAnyVoiceFew": "কয়েক সেকেন্ডের রেকর্ডিং থেকে যেকোনো কণ্ঠে পড়ে: সাথে থাকা একটি, বা বিনের নমুনা।",
"speech.readsVoiceHeardClip": "টাইমলাইনে নির্বাচিত ক্লিপে শোনা কণ্ঠে পড়ে",
"speech.readsVoiceHeardRecording": "বিনের রেকর্ডিংয়ে শোনা কণ্ঠে পড়ে। নিচে নির্বাচিত পয়েন্ট থেকে প্রথম দশ সেকেন্ড মডেলটি শোনে।",
diff --git a/src/crates/concat/locales/de.json b/src/crates/concat/locales/de.json
index f1aa103c..44f4208d 100644
--- a/src/crates/concat/locales/de.json
+++ b/src/crates/concat/locales/de.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Ein Titel pro gesprochenem Satz, im Takt der Wörter geschnitten, auf der ersten freien Spur.",
"captions.oneTitlePerSentence": "One title per sentence, held for as long as it takes to read, one after another on the first lane with room.",
"captions.placement": "Platzierung",
+ "captions.readingTheSound": "Ton des Clips wird gelesen",
"captions.transcribing": "Wird transkribiert",
"captions.typeWordsPutScreen": "Type the words to put on screen…",
"captions.useVoiceFromTimeline": "Stimme aus der Zeitleiste verwenden",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "In {0}",
"clipInspector.treatments": "Aufbereitung",
"clipInspector.volume": "Lautstärke",
+ "common.aboutLeft": "Noch etwa {0}:{1}",
"common.adjust": "Anpassen",
"common.apply": "Anwenden",
"common.audio": "Audio",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Importiere zuerst einen Clip, in dem eine Stimme zu hören ist.",
"speech.italian": "Italienisch",
"speech.japanese": "Japanisch",
+ "speech.loadingTheModel": "Stimmmodell wird geladen",
"speech.long": "lang",
"speech.longer": "länger",
"speech.male": "männlich",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Wähle eine Probe, in der eine Stimme zu hören ist",
"speech.portuguese": "Portugiesisch",
"speech.quick": "Schnell",
+ "speech.readingTheSample": "Stimmaufnahme wird gelesen",
"speech.readsAnyVoiceFew": "Liest in jeder Stimme aus ein paar Sekunden einer Aufnahme: einer, die mitgeliefert wird, oder einer Probe aus den Medien.",
"speech.readsVoiceHeardClip": "Liest in der Stimme, die im auf der Zeitleiste ausgewählten Clip zu hören ist",
"speech.readsVoiceHeardRecording": "Liest in der Stimme, die in einer Aufnahme aus den Medien zu hören ist. Die ersten zehn Sekunden ab der unten gewählten Stelle sind das, worauf das Modell hört.",
diff --git a/src/crates/concat/locales/en.json b/src/crates/concat/locales/en.json
index 973ef621..773b11f5 100644
--- a/src/crates/concat/locales/en.json
+++ b/src/crates/concat/locales/en.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "One title per phrase spoken, cut to the words' own timing, on the first lane with room.",
"captions.oneTitlePerSentence": "One title per sentence, held for as long as it takes to read, one after another on the first lane with room.",
"captions.placement": "Placement",
+ "captions.readingTheSound": "Reading the clip's sound",
"captions.transcribing": "Transcribing",
"captions.typeWordsPutScreen": "Type the words to put on screen…",
"captions.useVoiceFromTimeline": "Use voice from timeline",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "Into {0}",
"clipInspector.treatments": "Treatments",
"clipInspector.volume": "Volume",
+ "common.aboutLeft": "About {0}:{1} left",
"common.adjust": "Adjust",
"common.apply": "Apply",
"common.audio": "Audio",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Import a clip with a voice in it first.",
"speech.italian": "Italian",
"speech.japanese": "Japanese",
+ "speech.loadingTheModel": "Loading the voice model",
"speech.long": "long",
"speech.longer": "longer",
"speech.male": "male",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Pick a sample with a voice in it",
"speech.portuguese": "Portuguese",
"speech.quick": "Quick",
+ "speech.readingTheSample": "Reading the voice recording",
"speech.readsAnyVoiceFew": "Reads in any voice from a few seconds of a recording: one it ships with, or a sample from the bin.",
"speech.readsVoiceHeardClip": "Reads in the voice heard in the clip selected on the timeline",
"speech.readsVoiceHeardRecording": "Reads in the voice heard in a recording from the bin. The first ten seconds from the point chosen below are what the model listens to.",
diff --git a/src/crates/concat/locales/es.json b/src/crates/concat/locales/es.json
index da9c8bc2..5264fe7f 100644
--- a/src/crates/concat/locales/es.json
+++ b/src/crates/concat/locales/es.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Un título por frase hablada, ajustado al ritmo de las palabras, en la primera pista con espacio.",
"captions.oneTitlePerSentence": "One title per sentence, held for as long as it takes to read, one after another on the first lane with room.",
"captions.placement": "Posición",
+ "captions.readingTheSound": "Leyendo el sonido del clip",
"captions.transcribing": "Transcribiendo",
"captions.typeWordsPutScreen": "Type the words to put on screen…",
"captions.useVoiceFromTimeline": "Usar la voz de la línea de tiempo",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "Hacia {0}",
"clipInspector.treatments": "Tratamientos",
"clipInspector.volume": "Volumen",
+ "common.aboutLeft": "Quedan unos {0}:{1}",
"common.adjust": "Ajustar",
"common.apply": "Aplicar",
"common.audio": "Audio",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Importa primero un clip con una voz.",
"speech.italian": "Italiano",
"speech.japanese": "Japonés",
+ "speech.loadingTheModel": "Cargando el modelo de voz",
"speech.long": "larga",
"speech.longer": "más larga",
"speech.male": "masculina",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Elige una muestra con una voz",
"speech.portuguese": "Portugués",
"speech.quick": "Rápido",
+ "speech.readingTheSample": "Leyendo la grabación de voz",
"speech.readsAnyVoiceFew": "Lee con cualquier voz a partir de unos segundos de una grabación: una de las incluidas o una muestra de la biblioteca de medios.",
"speech.readsVoiceHeardClip": "Lee con la voz que se oye en el clip seleccionado en la línea de tiempo",
"speech.readsVoiceHeardRecording": "Lee con la voz que se oye en una grabación de la biblioteca de medios. El modelo escucha los diez primeros segundos desde el punto elegido abajo.",
diff --git a/src/crates/concat/locales/fa.json b/src/crates/concat/locales/fa.json
index 0b059a54..b0430f39 100644
--- a/src/crates/concat/locales/fa.json
+++ b/src/crates/concat/locales/fa.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "برای هر Phrase گفته‌شده یک عنوان ساخته می‌شود، با Timing خود کلمات، روی اولین لاین دارای فضای کافی.",
"captions.oneTitlePerSentence": "برای هر جمله یک عنوان ساخته می‌شود و به‌اندازه زمان لازم برای خواندن نمایش داده می‌شود؛ پشت سر هم روی اولین لاین دارای فضای کافی.",
"captions.placement": "جای‌گذاری",
+ "captions.readingTheSound": "در حال خواندن صدای کلیپ",
"captions.transcribing": "در حال رونویسی",
"captions.typeWordsPutScreen": "متنی را که باید روی صفحه نمایش داده شود وارد کنید…",
"captions.useVoiceFromTimeline": "استفاده از صدای خط زمان",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "به {0}",
"clipInspector.treatments": "پردازش‌ها",
"clipInspector.volume": "ولوم",
+ "common.aboutLeft": "حدود {0}:{1} مانده",
"common.adjust": "تنظیم",
"common.apply": "اعمال",
"common.audio": "صدا",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "ابتدا کلیپی را که صدایی در آن هست وارد کنید.",
"speech.italian": "ایتالیایی",
"speech.japanese": "ژاپنی",
+ "speech.loadingTheModel": "در حال بارگذاری مدل صدا",
"speech.long": "طولانی",
"speech.longer": "طولانی‌تر",
"speech.male": "مرد",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "نمونه‌ای را که صدایی در آن هست انتخاب کنید",
"speech.portuguese": "پرتغالی",
"speech.quick": "سریع",
+ "speech.readingTheSample": "در حال خواندن ضبط صدا",
"speech.readsAnyVoiceFew": "با هر صدایی از چند ثانیه ضبط می‌خواند: یکی از ضبط‌های همراه خودش، یا یک نمونه از Bin.",
"speech.readsVoiceHeardClip": "با صدایی که در کلیپ انتخاب‌شده روی تایم‌لاین شنیده می‌شود می‌خواند",
"speech.readsVoiceHeardRecording": "با صدایی که در یک ضبط از Bin شنیده می‌شود می‌خواند. مدل به ده ثانیه اول از نقطه انتخاب‌شده در پایین گوش می‌دهد.",
diff --git a/src/crates/concat/locales/fr.json b/src/crates/concat/locales/fr.json
index 8a0b82f3..cc18ea3a 100644
--- a/src/crates/concat/locales/fr.json
+++ b/src/crates/concat/locales/fr.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Un titre par phrase prononcée, calé sur le rythme des mots, sur la première piste libre.",
"captions.oneTitlePerSentence": "One title per sentence, held for as long as it takes to read, one after another on the first lane with room.",
"captions.placement": "Position",
+ "captions.readingTheSound": "Lecture du son du clip",
"captions.transcribing": "Transcription en cours",
"captions.typeWordsPutScreen": "Type the words to put on screen…",
"captions.useVoiceFromTimeline": "Utiliser la voix de la timeline",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "Vers {0}",
"clipInspector.treatments": "Traitements",
"clipInspector.volume": "Volume",
+ "common.aboutLeft": "Encore environ {0}:{1}",
"common.adjust": "Réglages",
"common.apply": "Appliquer",
"common.audio": "Audio",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Importez d'abord un clip contenant une voix.",
"speech.italian": "Italien",
"speech.japanese": "Japonais",
+ "speech.loadingTheModel": "Chargement du modèle de voix",
"speech.long": "longues",
"speech.longer": "plus longues",
"speech.male": "masculine",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Choisissez un échantillon contenant une voix",
"speech.portuguese": "Portugais",
"speech.quick": "Rapide",
+ "speech.readingTheSample": "Lecture de l'enregistrement de voix",
"speech.readsAnyVoiceFew": "Lit avec n'importe quelle voix à partir de quelques secondes d'un enregistrement : l'un de ceux qu'il embarque, ou un échantillon du chutier.",
"speech.readsVoiceHeardClip": "Lit avec la voix entendue dans le clip sélectionné sur la timeline",
"speech.readsVoiceHeardRecording": "Lit avec la voix entendue dans un enregistrement du chutier. Les dix premières secondes à partir du point choisi ci-dessous sont ce que le modèle écoute.",
diff --git a/src/crates/concat/locales/he.json b/src/crates/concat/locales/he.json
index b23a4954..d087c5b1 100644
--- a/src/crates/concat/locales/he.json
+++ b/src/crates/concat/locales/he.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "כותרת אחת לכל ביטוי שנאמר, חתוכה לפי תזמון המילים עצמן, בערוץ הראשון שיש בו מקום.",
"captions.oneTitlePerSentence": "כותרת אחת לכל משפט, מוצגת כל עוד צריך כדי לקרוא אותה, בזו אחר זו בערוץ הראשון שיש בו מקום.",
"captions.placement": "מיקום",
+ "captions.readingTheSound": "קורא את הצליל של הקליפ",
"captions.transcribing": "מתמלל",
"captions.typeWordsPutScreen": "יש להקליד את המילים שיופיעו על המסך…",
"captions.useVoiceFromTimeline": "שימוש בקול מציר הזמן",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "אל {0}",
"clipInspector.treatments": "טיפולים",
"clipInspector.volume": "עוצמת קול",
+ "common.aboutLeft": "נותרו כ-{0}:{1}",
"common.adjust": "כוונון",
"common.apply": "החלה",
"common.audio": "שמע",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "יש לייבא קודם קליפ שיש בו קול.",
"speech.italian": "איטלקית",
"speech.japanese": "יפנית",
+ "speech.loadingTheModel": "טוען את מודל הקול",
"speech.long": "ארוכות",
"speech.longer": "ארוכות יותר",
"speech.male": "זכר",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "בחירת דגימה שיש בה קול",
"speech.portuguese": "פורטוגזית",
"speech.quick": "מהיר",
+ "speech.readingTheSample": "קורא את הקלטת הקול",
"speech.readsAnyVoiceFew": "קורא בכל קול מכמה שניות של הקלטה: אחת שמגיעה איתו, או דגימה מהמאגר.",
"speech.readsVoiceHeardClip": "קורא בקול שנשמע בקליפ שנבחר בציר הזמן",
"speech.readsVoiceHeardRecording": "קורא בקול שנשמע בהקלטה מהמאגר. עשר השניות הראשונות מהנקודה שנבחרה למטה הן מה שהמודל מקשיב לו.",
diff --git a/src/crates/concat/locales/hi.json b/src/crates/concat/locales/hi.json
index 92fe0a0e..82d36721 100644
--- a/src/crates/concat/locales/hi.json
+++ b/src/crates/concat/locales/hi.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "बोले गए हर वाक्यांश के लिए एक टाइटल, शब्दों की अपनी टाइमिंग पर कटा हुआ, खाली जगह वाली पहली लेन पर।",
"captions.oneTitlePerSentence": "हर वाक्य के लिए एक टाइटल, जितनी देर पढ़ने में लगे उतनी देर तक, एक के बाद एक, खाली जगह वाली पहली लेन पर।",
"captions.placement": "स्थान",
+ "captions.readingTheSound": "क्लिप की आवाज़ पढ़ी जा रही है",
"captions.transcribing": "ट्रांसक्राइब हो रहा है",
"captions.typeWordsPutScreen": "स्क्रीन पर दिखाने के लिए शब्द लिखें…",
"captions.useVoiceFromTimeline": "टाइमलाइन की आवाज़ इस्तेमाल करें",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "{0} में",
"clipInspector.treatments": "ऑडियो सुधार",
"clipInspector.volume": "वॉल्यूम",
+ "common.aboutLeft": "लगभग {0}:{1} बाकी",
"common.adjust": "समायोजन",
"common.apply": "लागू करें",
"common.audio": "ऑडियो",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "पहले आवाज़ वाली कोई क्लिप इंपोर्ट करें।",
"speech.italian": "इतालवी",
"speech.japanese": "जापानी",
+ "speech.loadingTheModel": "आवाज़ मॉडल लोड हो रहा है",
"speech.long": "लंबे",
"speech.longer": "और लंबे",
"speech.male": "पुरुष",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "आवाज़ वाला कोई नमूना चुनें",
"speech.portuguese": "पुर्तगाली",
"speech.quick": "तेज़",
+ "speech.readingTheSample": "आवाज़ की रिकॉर्डिंग पढ़ी जा रही है",
"speech.readsAnyVoiceFew": "रिकॉर्डिंग के कुछ सेकंड से किसी भी आवाज़ में पढ़ता है: साथ आई किसी रिकॉर्डिंग से, या बिन के किसी नमूने से।",
"speech.readsVoiceHeardClip": "टाइमलाइन पर चुनी गई क्लिप में सुनाई देने वाली आवाज़ में पढ़ता है",
"speech.readsVoiceHeardRecording": "बिन की किसी रिकॉर्डिंग में सुनाई देने वाली आवाज़ में पढ़ता है। नीचे चुने गए बिंदु से पहले दस सेकंड वह हिस्सा हैं जिसे मॉडल सुनता है।",
diff --git a/src/crates/concat/locales/hr.json b/src/crates/concat/locales/hr.json
index ae2f6e52..f825f515 100644
--- a/src/crates/concat/locales/hr.json
+++ b/src/crates/concat/locales/hr.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Jedan naslov po izgovorenoj frazi, rezan na vlastiti tajming riječi, na prvoj traci s mjesta.",
"captions.oneTitlePerSentence": "Jedan naslov po rečenici, držan dok se ne pročita, jedan za drugim na prvoj traci s mjesta.",
"captions.placement": "Smještaj",
+ "captions.readingTheSound": "Čitanje zvuka isječka",
"captions.transcribing": "Transkripcija",
"captions.typeWordsPutScreen": "Upišite riječi za prikaz na zaslonu…",
"captions.useVoiceFromTimeline": "Koristi glas s vremenske crte",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "U {0}",
"clipInspector.treatments": "Tretmani",
"clipInspector.volume": "Glasnoća",
+ "common.aboutLeft": "Još oko {0}:{1}",
"common.adjust": "Prilagodi",
"common.apply": "Primijeni",
"common.audio": "Zvuk",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Prvo uvezite klip u kojem je glas.",
"speech.italian": "talijanski",
"speech.japanese": "japanski",
+ "speech.loadingTheModel": "Učitavanje modela glasa",
"speech.long": "dugo",
"speech.longer": "duže",
"speech.male": "muški",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Odaberite uzorak u kojem je glas",
"speech.portuguese": "portugalski",
"speech.quick": "Brzo",
+ "speech.readingTheSample": "Čitanje snimke glasa",
"speech.readsAnyVoiceFew": "Čita bilo kojim glasom iz nekoliko sekundi snimke: one koja dolazi s njim ili uzorka iz spremnika.",
"speech.readsVoiceHeardClip": "Čita glasom koji se čuje u klipu odabranom na vremenskoj liniji",
"speech.readsVoiceHeardRecording": "Čita glasom koji se čuje u snimci iz spremnika. Model sluša prvih deset sekundi od točke odabrane ispod.",
diff --git a/src/crates/concat/locales/id.json b/src/crates/concat/locales/id.json
index 3ace0c43..1eccfb15 100644
--- a/src/crates/concat/locales/id.json
+++ b/src/crates/concat/locales/id.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Satu judul untuk setiap frasa yang diucapkan, dipotong mengikuti waktu kata-katanya, di lajur pertama yang masih lowong.",
"captions.oneTitlePerSentence": "Satu judul untuk setiap kalimat, ditampilkan selama waktu yang dibutuhkan untuk membacanya, berurutan di lajur pertama yang masih lowong.",
"captions.placement": "Penempatan",
+ "captions.readingTheSound": "Membaca suara klip",
"captions.transcribing": "Mentranskripsi",
"captions.typeWordsPutScreen": "Ketik kata-kata yang akan tampil di layar…",
"captions.useVoiceFromTimeline": "Gunakan suara dari linimasa",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "Ke {0}",
"clipInspector.treatments": "Perbaikan audio",
"clipInspector.volume": "Volume",
+ "common.aboutLeft": "Sekitar {0}:{1} lagi",
"common.adjust": "Sesuaikan",
"common.apply": "Terapkan",
"common.audio": "Audio",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Impor dulu klip yang berisi suara.",
"speech.italian": "Italia",
"speech.japanese": "Jepang",
+ "speech.loadingTheModel": "Memuat model suara",
"speech.long": "panjang",
"speech.longer": "lebih panjang",
"speech.male": "laki-laki",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Pilih sampel yang berisi suara",
"speech.portuguese": "Portugis",
"speech.quick": "Cepat",
+ "speech.readingTheSample": "Membaca rekaman suara",
"speech.readsAnyVoiceFew": "Membaca dengan suara apa pun dari beberapa detik rekaman: rekaman bawaannya, atau sampel dari bin.",
"speech.readsVoiceHeardClip": "Membaca dengan suara yang terdengar di klip yang dipilih di linimasa",
"speech.readsVoiceHeardRecording": "Membaca dengan suara yang terdengar di rekaman dari bin. Sepuluh detik pertama dari titik yang dipilih di bawah adalah bagian yang didengarkan model.",
diff --git a/src/crates/concat/locales/it.json b/src/crates/concat/locales/it.json
index 0161acc2..76ba8004 100644
--- a/src/crates/concat/locales/it.json
+++ b/src/crates/concat/locales/it.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Un titolo per ogni frase pronunciata, tagliato sul ritmo delle parole, sulla prima traccia libera.",
"captions.oneTitlePerSentence": "One title per sentence, held for as long as it takes to read, one after another on the first lane with room.",
"captions.placement": "Posizione",
+ "captions.readingTheSound": "Lettura del suono della clip",
"captions.transcribing": "Trascrizione in corso",
"captions.typeWordsPutScreen": "Type the words to put on screen…",
"captions.useVoiceFromTimeline": "Usa la voce dalla timeline",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "Verso {0}",
"clipInspector.treatments": "Trattamenti",
"clipInspector.volume": "Volume",
+ "common.aboutLeft": "Mancano circa {0}:{1}",
"common.adjust": "Regola",
"common.apply": "Applica",
"common.audio": "Audio",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Importa prima una clip con una voce.",
"speech.italian": "Italiano",
"speech.japanese": "Giapponese",
+ "speech.loadingTheModel": "Caricamento del modello vocale",
"speech.long": "lunghe",
"speech.longer": "più lunghe",
"speech.male": "maschile",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Scegli un campione con una voce",
"speech.portuguese": "Portoghese",
"speech.quick": "Rapida",
+ "speech.readingTheSample": "Lettura della registrazione vocale",
"speech.readsAnyVoiceFew": "Legge con qualsiasi voce a partire da pochi secondi di una registrazione: una fornita con il modello, o un campione dal pannello Media.",
"speech.readsVoiceHeardClip": "Legge con la voce che si sente nella clip selezionata sulla timeline",
"speech.readsVoiceHeardRecording": "Legge con la voce che si sente in una registrazione del pannello Media. Il modello ascolta i primi dieci secondi dal punto scelto qui sotto.",
diff --git a/src/crates/concat/locales/ja.json b/src/crates/concat/locales/ja.json
index 11da7af1..2a83f8c1 100644
--- a/src/crates/concat/locales/ja.json
+++ b/src/crates/concat/locales/ja.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "発話されたフレーズごとにタイトルを一つ、言葉のタイミングどおりに、空いている最初のレーンへ。",
"captions.oneTitlePerSentence": "一文ごとにタイトルを一つ、読み終えるのに必要な時間だけ表示し、空いている最初のレーンに順番に並べます。",
"captions.placement": "配置",
+ "captions.readingTheSound": "クリップの音声を読み込み中",
"captions.transcribing": "文字起こし中",
"captions.typeWordsPutScreen": "画面に表示する言葉を入力…",
"captions.useVoiceFromTimeline": "タイムラインの音声を使う",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "{0} へ",
"clipInspector.treatments": "処理",
"clipInspector.volume": "音量",
+ "common.aboutLeft": "残り約 {0}:{1}",
"common.adjust": "調整",
"common.apply": "適用",
"common.audio": "オーディオ",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "まず声の入ったクリップを読み込んでください。",
"speech.italian": "イタリア語",
"speech.japanese": "日本語",
+ "speech.loadingTheModel": "音声モデルを読み込み中",
"speech.long": "長い",
"speech.longer": "より長い",
"speech.male": "男性",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "声の入ったサンプルを選んでください",
"speech.portuguese": "ポルトガル語",
"speech.quick": "クイック",
+ "speech.readingTheSample": "声の録音を読み込み中",
"speech.readsAnyVoiceFew": "数秒の録音からどんな声でも読み上げます : 付属の録音か、ビンのサンプルから。",
"speech.readsVoiceHeardClip": "タイムラインで選択したクリップに聞こえる声で読み上げます",
"speech.readsVoiceHeardRecording": "ビンの録音に聞こえる声で読み上げます。モデルが聞くのは、下で選んだ位置から最初の 10 秒です。",
diff --git a/src/crates/concat/locales/ko.json b/src/crates/concat/locales/ko.json
index db9be476..b93f7b57 100644
--- a/src/crates/concat/locales/ko.json
+++ b/src/crates/concat/locales/ko.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "말한 구절마다 자막 하나씩, 말의 타이밍에 맞춰, 자리가 있는 첫 레인에 놓습니다.",
"captions.oneTitlePerSentence": "One title per sentence, held for as long as it takes to read, one after another on the first lane with room.",
"captions.placement": "배치",
+ "captions.readingTheSound": "클립 소리 읽는 중",
"captions.transcribing": "받아쓰는 중",
"captions.typeWordsPutScreen": "Type the words to put on screen…",
"captions.useVoiceFromTimeline": "타임라인의 음성 사용",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "{0}(으)로",
"clipInspector.treatments": "처리",
"clipInspector.volume": "볼륨",
+ "common.aboutLeft": "약 {0}:{1} 남음",
"common.adjust": "조정",
"common.apply": "적용",
"common.audio": "오디오",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "먼저 목소리가 담긴 클립을 가져오세요.",
"speech.italian": "이탈리아어",
"speech.japanese": "일본어",
+ "speech.loadingTheModel": "음성 모델 불러오는 중",
"speech.long": "길게",
"speech.longer": "더 길게",
"speech.male": "남성",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "목소리가 담긴 샘플을 선택하세요",
"speech.portuguese": "포르투갈어",
"speech.quick": "빠름",
+ "speech.readingTheSample": "목소리 녹음 읽는 중",
"speech.readsAnyVoiceFew": "녹음 몇 초만 있으면 어떤 목소리로든 읽습니다: 딸려 오는 녹음이나 미디어 빈의 샘플.",
"speech.readsVoiceHeardClip": "타임라인에서 선택한 클립 속 목소리로 읽습니다",
"speech.readsVoiceHeardRecording": "미디어 빈에 있는 녹음 속 목소리로 읽습니다. 아래에서 고른 지점부터 10초 동안을 모델이 듣습니다.",
diff --git a/src/crates/concat/locales/nl.json b/src/crates/concat/locales/nl.json
index 5e6b663e..6d242dc8 100644
--- a/src/crates/concat/locales/nl.json
+++ b/src/crates/concat/locales/nl.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Eén titel per gesproken zin, getimed op de woorden zelf, op het eerste spoor met ruimte.",
"captions.oneTitlePerSentence": "Eén titel per zin, zo lang in beeld als nodig is om hem te lezen, na elkaar op het eerste spoor met ruimte.",
"captions.placement": "Plaatsing",
+ "captions.readingTheSound": "Geluid van de clip lezen",
"captions.transcribing": "Transcriberen",
"captions.typeWordsPutScreen": "Typ de woorden die in beeld moeten komen…",
"captions.useVoiceFromTimeline": "Stem uit de tijdlijn gebruiken",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "Naar {0}",
"clipInspector.treatments": "Bewerkingen",
"clipInspector.volume": "Volume",
+ "common.aboutLeft": "Nog ongeveer {0}:{1}",
"common.adjust": "Aanpassen",
"common.apply": "Toepassen",
"common.audio": "Audio",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Importeer eerst een clip waarin een stem te horen is.",
"speech.italian": "Italiaans",
"speech.japanese": "Japans",
+ "speech.loadingTheModel": "Stemmodel laden",
"speech.long": "lang",
"speech.longer": "langer",
"speech.male": "man",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Kies een sample met een stem erin",
"speech.portuguese": "Portugees",
"speech.quick": "Snel",
+ "speech.readingTheSample": "Stemopname lezen",
"speech.readsAnyVoiceFew": "Leest met elke stem op basis van een paar seconden opname: een meegeleverde, of een sample uit de mediabak.",
"speech.readsVoiceHeardClip": "Leest met de stem die te horen is in de geselecteerde clip op de tijdlijn",
"speech.readsVoiceHeardRecording": "Leest met de stem die te horen is in een opname uit de mediabak. Het model luistert naar de eerste tien seconden vanaf het hieronder gekozen punt.",
diff --git a/src/crates/concat/locales/pl.json b/src/crates/concat/locales/pl.json
index 4c9a75e9..b0aa9255 100644
--- a/src/crates/concat/locales/pl.json
+++ b/src/crates/concat/locales/pl.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Jeden napis na każdą wypowiedzianą frazę, przycięty do rytmu samych słów, na pierwszej ścieżce z wolnym miejscem.",
"captions.oneTitlePerSentence": "Jeden napis na zdanie, widoczny tak długo, jak trzeba, by go przeczytać, jeden po drugim na pierwszej ścieżce z wolnym miejscem.",
"captions.placement": "Położenie",
+ "captions.readingTheSound": "Odczytywanie dźwięku klipu",
"captions.transcribing": "Transkrypcja",
"captions.typeWordsPutScreen": "Wpisz słowa, które pojawią się na ekranie…",
"captions.useVoiceFromTimeline": "Użyj głosu z osi czasu",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "W: {0}",
"clipInspector.treatments": "Obróbka",
"clipInspector.volume": "Głośność",
+ "common.aboutLeft": "Zostało około {0}:{1}",
"common.adjust": "Korekta",
"common.apply": "Zastosuj",
"common.audio": "Audio",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Najpierw zaimportuj klip z głosem.",
"speech.italian": "włoski",
"speech.japanese": "japoński",
+ "speech.loadingTheModel": "Ładowanie modelu głosu",
"speech.long": "długie",
"speech.longer": "dłuższe",
"speech.male": "mężczyzna",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Wybierz próbkę z głosem",
"speech.portuguese": "portugalski",
"speech.quick": "Szybko",
+ "speech.readingTheSample": "Odczytywanie nagrania głosu",
"speech.readsAnyVoiceFew": "Czyta dowolnym głosem na podstawie kilku sekund nagrania: dołączonego albo próbki z pojemnika.",
"speech.readsVoiceHeardClip": "Czyta głosem słyszanym w klipie zaznaczonym na osi czasu",
"speech.readsVoiceHeardRecording": "Czyta głosem słyszanym w nagraniu z pojemnika. Model słucha pierwszych dziesięciu sekund od wybranego poniżej miejsca.",
diff --git a/src/crates/concat/locales/pt-BR.json b/src/crates/concat/locales/pt-BR.json
index 5ae6e433..d890b1db 100644
--- a/src/crates/concat/locales/pt-BR.json
+++ b/src/crates/concat/locales/pt-BR.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Um título por frase falada, no ritmo das palavras, na primeira faixa com espaço.",
"captions.oneTitlePerSentence": "One title per sentence, held for as long as it takes to read, one after another on the first lane with room.",
"captions.placement": "Posição",
+ "captions.readingTheSound": "Lendo o som do clipe",
"captions.transcribing": "Transcrevendo",
"captions.typeWordsPutScreen": "Type the words to put on screen…",
"captions.useVoiceFromTimeline": "Usar a voz da linha do tempo",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "Para {0}",
"clipInspector.treatments": "Tratamentos",
"clipInspector.volume": "Volume",
+ "common.aboutLeft": "Faltam cerca de {0}:{1}",
"common.adjust": "Ajustar",
"common.apply": "Aplicar",
"common.audio": "Áudio",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Importe primeiro um clipe com uma voz.",
"speech.italian": "Italiano",
"speech.japanese": "Japonês",
+ "speech.loadingTheModel": "Carregando o modelo de voz",
"speech.long": "longo",
"speech.longer": "mais longo",
"speech.male": "masculina",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Escolha uma amostra com uma voz",
"speech.portuguese": "Português",
"speech.quick": "Rápido",
+ "speech.readingTheSample": "Lendo a gravação de voz",
"speech.readsAnyVoiceFew": "Lê em qualquer voz a partir de alguns segundos de uma gravação: uma que vem com ele, ou uma amostra da biblioteca.",
"speech.readsVoiceHeardClip": "Lê na voz ouvida no clipe selecionado na linha do tempo",
"speech.readsVoiceHeardRecording": "Lê na voz ouvida numa gravação da biblioteca. Os primeiros dez segundos a partir do ponto escolhido abaixo são o que o modelo escuta.",
diff --git a/src/crates/concat/locales/ru.json b/src/crates/concat/locales/ru.json
index bb5ac03c..8d366535 100644
--- a/src/crates/concat/locales/ru.json
+++ b/src/crates/concat/locales/ru.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Один титр на каждую произнесённую фразу, в такт словам, на первой свободной дорожке.",
"captions.oneTitlePerSentence": "Один титр на предложение, держится столько, сколько нужно, чтобы прочитать, один за другим на первой свободной дорожке.",
"captions.placement": "Расположение",
+ "captions.readingTheSound": "Чтение звука клипа",
"captions.transcribing": "Распознавание",
"captions.typeWordsPutScreen": "Введите слова, которые появятся на экране…",
"captions.useVoiceFromTimeline": "Использовать голос с таймлайна",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "В {0}",
"clipInspector.treatments": "Обработка",
"clipInspector.volume": "Громкость",
+ "common.aboutLeft": "Осталось около {0}:{1}",
"common.adjust": "Коррекция",
"common.apply": "Применить",
"common.audio": "Аудио",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Сначала импортируйте клип с голосом.",
"speech.italian": "Итальянский",
"speech.japanese": "Японский",
+ "speech.loadingTheModel": "Загрузка модели голоса",
"speech.long": "длинные",
"speech.longer": "ещё длиннее",
"speech.male": "мужской",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Выберите образец с голосом",
"speech.portuguese": "Португальский",
"speech.quick": "Быстрое",
+ "speech.readingTheSample": "Чтение записи голоса",
"speech.readsAnyVoiceFew": "Читает любым голосом по нескольким секундам записи: из комплекта или из образца в медиа.",
"speech.readsVoiceHeardClip": "Читает голосом, звучащим в клипе, выбранном на таймлайне",
"speech.readsVoiceHeardRecording": "Читает голосом, звучащим в записи из медиа. Модель слушает первые десять секунд от точки, выбранной ниже.",
diff --git a/src/crates/concat/locales/th.json b/src/crates/concat/locales/th.json
index 8bab5410..18b0df3b 100644
--- a/src/crates/concat/locales/th.json
+++ b/src/crates/concat/locales/th.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "ไตเติลหนึ่งรายการต่อหนึ่งวลีที่พูด ตัดตามจังหวะของคำ วางบนเลนแรกที่มีที่ว่าง",
"captions.oneTitlePerSentence": "ไตเติลหนึ่งรายการต่อหนึ่งประโยค แสดงค้างไว้นานพอให้อ่านทัน เรียงต่อกันบนเลนแรกที่มีที่ว่าง",
"captions.placement": "ตำแหน่งที่วาง",
+ "captions.readingTheSound": "กำลังอ่านเสียงของคลิป",
"captions.transcribing": "กำลังถอดเสียง",
"captions.typeWordsPutScreen": "พิมพ์ข้อความที่จะแสดงบนจอ…",
"captions.useVoiceFromTimeline": "ใช้เสียงพูดจากไทม์ไลน์",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "เข้าสู่ {0}",
"clipInspector.treatments": "การปรับปรุงเสียง",
"clipInspector.volume": "ระดับเสียง",
+ "common.aboutLeft": "เหลืออีกประมาณ {0}:{1}",
"common.adjust": "ปรับแต่ง",
"common.apply": "นำไปใช้",
"common.audio": "เสียง",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "นำเข้าคลิปที่มีเสียงพูดก่อน",
"speech.italian": "อิตาลี",
"speech.japanese": "ญี่ปุ่น",
+ "speech.loadingTheModel": "กำลังโหลดโมเดลเสียง",
"speech.long": "ยาว",
"speech.longer": "ยาวขึ้น",
"speech.male": "ชาย",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "เลือกตัวอย่างที่มีเสียงพูด",
"speech.portuguese": "โปรตุเกส",
"speech.quick": "เร็ว",
+ "speech.readingTheSample": "กำลังอ่านไฟล์บันทึกเสียง",
"speech.readsAnyVoiceFew": "อ่านด้วยเสียงใดก็ได้จากไฟล์บันทึกเพียงไม่กี่วินาที: ไฟล์ที่มาพร้อมกัน หรือตัวอย่างจากคลังสื่อ",
"speech.readsVoiceHeardClip": "อ่านด้วยเสียงที่ได้ยินในคลิปที่เลือกบนไทม์ไลน์",
"speech.readsVoiceHeardRecording": "อ่านด้วยเสียงที่ได้ยินในไฟล์บันทึกจากคลังสื่อ โมเดลจะฟังสิบวินาทีแรกนับจากจุดที่เลือกด้านล่าง",
diff --git a/src/crates/concat/locales/tr.json b/src/crates/concat/locales/tr.json
index 7dbc9757..de7062bf 100644
--- a/src/crates/concat/locales/tr.json
+++ b/src/crates/concat/locales/tr.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Söylenen her cümle için bir yazı, kelimelerin kendi zamanlamasıyla, yeri olan ilk şeritte.",
"captions.oneTitlePerSentence": "One title per sentence, held for as long as it takes to read, one after another on the first lane with room.",
"captions.placement": "Yerleşim",
+ "captions.readingTheSound": "Klibin sesi okunuyor",
"captions.transcribing": "Yazıya dökülüyor",
"captions.typeWordsPutScreen": "Type the words to put on screen…",
"captions.useVoiceFromTimeline": "Zaman çizelgesindeki sesi kullan",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "{0} klibine",
"clipInspector.treatments": "İşlemler",
"clipInspector.volume": "Ses düzeyi",
+ "common.aboutLeft": "Yaklaşık {0}:{1} kaldı",
"common.adjust": "Ayarla",
"common.apply": "Uygula",
"common.audio": "Ses",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Önce içinde ses olan bir klip içe aktarın.",
"speech.italian": "İtalyanca",
"speech.japanese": "Japonca",
+ "speech.loadingTheModel": "Ses modeli yükleniyor",
"speech.long": "uzun",
"speech.longer": "daha uzun",
"speech.male": "erkek",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "İçinde ses olan bir örnek seçin",
"speech.portuguese": "Portekizce",
"speech.quick": "Hızlı",
+ "speech.readingTheSample": "Ses kaydı okunuyor",
"speech.readsAnyVoiceFew": "Bir kaydın birkaç saniyesinden herhangi bir sesle okur: beraberinde gelen bir kayıt ya da medya kutusundan bir örnek.",
"speech.readsVoiceHeardClip": "Zaman çizelgesinde seçili klipte duyulan sesle okur",
"speech.readsVoiceHeardRecording": "Medya kutusundaki bir kayıtta duyulan sesle okur. Model, aşağıda seçilen noktadan itibaren ilk on saniyeyi dinler.",
diff --git a/src/crates/concat/locales/uk.json b/src/crates/concat/locales/uk.json
index 103293ce..94c9a1c3 100644
--- a/src/crates/concat/locales/uk.json
+++ b/src/crates/concat/locales/uk.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Один титр на кожну вимовлену фразу, точно за таймінгом слів, на першій вільній доріжці.",
"captions.oneTitlePerSentence": "Один титр на речення, рівно стільки, скільки треба, щоб його прочитати, один за одним на першій вільній доріжці.",
"captions.placement": "Розташування",
+ "captions.readingTheSound": "Читання звуку кліпу",
"captions.transcribing": "Розпізнавання",
"captions.typeWordsPutScreen": "Введіть слова, які з’являться на екрані…",
"captions.useVoiceFromTimeline": "Використати голос із часової шкали",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "У {0}",
"clipInspector.treatments": "Обробка",
"clipInspector.volume": "Гучність",
+ "common.aboutLeft": "Залишилось близько {0}:{1}",
"common.adjust": "Корекція",
"common.apply": "Застосувати",
"common.audio": "Аудіо",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Спершу імпортуйте кліп із голосом.",
"speech.italian": "італійський",
"speech.japanese": "японський",
+ "speech.loadingTheModel": "Завантаження моделі голосу",
"speech.long": "довгі",
"speech.longer": "довші",
"speech.male": "чоловічий",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Виберіть зразок із голосом",
"speech.portuguese": "португальський",
"speech.quick": "Швидко",
+ "speech.readingTheSample": "Читання запису голосу",
"speech.readsAnyVoiceFew": "Читає будь-яким голосом за кількома секундами запису: вбудованим або зразком із медіатеки.",
"speech.readsVoiceHeardClip": "Читає голосом, який звучить у вибраному на часовій шкалі кліпі",
"speech.readsVoiceHeardRecording": "Читає голосом, який звучить у записі з медіатеки. Модель слухає перші десять секунд від вибраної нижче точки.",
diff --git a/src/crates/concat/locales/ur.json b/src/crates/concat/locales/ur.json
index 3f02197c..047ff593 100644
--- a/src/crates/concat/locales/ur.json
+++ b/src/crates/concat/locales/ur.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "ہر بولے گئے فقرے کے لیے ایک ٹائٹل، الفاظ کی اپنی ٹائمنگ کے مطابق کٹا ہوا، پہلی ایسی لین پر جس میں جگہ ہو۔",
"captions.oneTitlePerSentence": "ہر جملے کے لیے ایک ٹائٹل، جو پڑھنے میں لگنے والے وقت تک رہتا ہے، ایک کے بعد ایک پہلی ایسی لین پر جس میں جگہ ہو۔",
"captions.placement": "جگہ",
+ "captions.readingTheSound": "کلپ کی آواز پڑھی جا رہی ہے",
"captions.transcribing": "ٹرانسکرائب ہو رہا ہے",
"captions.typeWordsPutScreen": "اسکرین پر دکھانے کے لیے الفاظ لکھیں…",
"captions.useVoiceFromTimeline": "ٹائم لائن کی آواز استعمال کریں",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "{0} میں",
"clipInspector.treatments": "علاج",
"clipInspector.volume": "والیوم",
+ "common.aboutLeft": "تقریباً {0}:{1} باقی",
"common.adjust": "ایڈجسٹ",
"common.apply": "لاگو کریں",
"common.audio": "آڈیو",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "پہلے کوئی ایسا کلپ امپورٹ کریں جس میں آواز ہو۔",
"speech.italian": "اطالوی",
"speech.japanese": "جاپانی",
+ "speech.loadingTheModel": "آواز کا ماڈل لوڈ ہو رہا ہے",
"speech.long": "لمبے",
"speech.longer": "زیادہ لمبے",
"speech.male": "مرد",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "کوئی ایسا نمونہ چنیں جس میں آواز ہو",
"speech.portuguese": "پرتگالی",
"speech.quick": "تیز",
+ "speech.readingTheSample": "آواز کی ریکارڈنگ پڑھی جا رہی ہے",
"speech.readsAnyVoiceFew": "کسی ریکارڈنگ کے چند سیکنڈ سے کسی بھی آواز میں پڑھتا ہے: ساتھ آنے والی کوئی ریکارڈنگ، یا بِن کا کوئی نمونہ۔",
"speech.readsVoiceHeardClip": "ٹائم لائن پر منتخب کلپ میں سنائی دینے والی آواز میں پڑھتا ہے",
"speech.readsVoiceHeardRecording": "بِن کی کسی ریکارڈنگ میں سنائی دینے والی آواز میں پڑھتا ہے۔ نیچے چنے گئے مقام سے پہلے دس سیکنڈ وہ ہیں جو ماڈل سنتا ہے۔",
diff --git a/src/crates/concat/locales/vi.json b/src/crates/concat/locales/vi.json
index ac85e620..bbe24d6a 100644
--- a/src/crates/concat/locales/vi.json
+++ b/src/crates/concat/locales/vi.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "Tạo một tiêu đề cho mỗi cụm từ được nói, căn theo thời điểm của từng từ và đặt trên track đầu tiên còn chỗ.",
"captions.oneTitlePerSentence": "Tạo một tiêu đề cho mỗi câu, giữ đủ lâu để người xem đọc, lần lượt trên track đầu tiên còn chỗ.",
"captions.placement": "Vị trí",
+ "captions.readingTheSound": "Đang đọc âm thanh của clip",
"captions.transcribing": "Đang chuyển đổi",
"captions.typeWordsPutScreen": "Nhập các từ muốn hiển thị trên màn hình…",
"captions.useVoiceFromTimeline": "Dùng giọng nói từ dòng thời gian",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "Sang {0}",
"clipInspector.treatments": "Xử lý hình ảnh",
"clipInspector.volume": "Âm lượng",
+ "common.aboutLeft": "Còn khoảng {0}:{1}",
"common.adjust": "Điều chỉnh",
"common.apply": "Áp dụng",
"common.audio": "Âm thanh",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "Trước tiên, hãy nhập clip có giọng nói.",
"speech.italian": "Ý",
"speech.japanese": "Nhật Bản",
+ "speech.loadingTheModel": "Đang tải mô hình giọng nói",
"speech.long": "dài",
"speech.longer": "dài hơn",
"speech.male": "nam",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "Chọn mẫu có giọng nói",
"speech.portuguese": "Bồ Đào Nha",
"speech.quick": "Nhanh",
+ "speech.readingTheSample": "Đang đọc bản ghi giọng nói",
"speech.readsAnyVoiceFew": "Có thể đọc bằng bất kỳ giọng nào từ vài giây ghi âm: bản ghi đi kèm hoặc mẫu trong thư viện.",
"speech.readsVoiceHeardClip": "Đọc bằng giọng nghe được trong clip đã chọn trên dòng thời gian",
"speech.readsVoiceHeardRecording": "Đọc bằng giọng nghe được trong bản ghi âm ở thư viện. Mô hình sẽ nghe mười giây đầu tiên tính từ vị trí được chọn bên dưới.",
diff --git a/src/crates/concat/locales/zh-Hans.json b/src/crates/concat/locales/zh-Hans.json
index ee442873..c05cf6b1 100644
--- a/src/crates/concat/locales/zh-Hans.json
+++ b/src/crates/concat/locales/zh-Hans.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "每个口语短语一个标题,按词语本身的时间切分,放在第一条有空位的轨道上。",
"captions.oneTitlePerSentence": "每句一个标题,停留足够阅读的时长,依次排列在第一条有空位的轨道上。",
"captions.placement": "位置",
+ "captions.readingTheSound": "正在读取片段的声音",
"captions.transcribing": "正在转录",
"captions.typeWordsPutScreen": "输入要显示在屏幕上的文字…",
"captions.useVoiceFromTimeline": "使用时间线中的语音",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "进入 {0}",
"clipInspector.treatments": "处理方式",
"clipInspector.volume": "音量",
+ "common.aboutLeft": "还剩约 {0}:{1}",
"common.adjust": "调整",
"common.apply": "应用",
"common.audio": "音频",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "请先导入一个含有人声的片段。",
"speech.italian": "意大利语",
"speech.japanese": "日语",
+ "speech.loadingTheModel": "正在加载语音模型",
"speech.long": "长",
"speech.longer": "更长",
"speech.male": "男声",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "选择一段含有人声的样本",
"speech.portuguese": "葡萄牙语",
"speech.quick": "快速",
+ "speech.readingTheSample": "正在读取声音录音",
"speech.readsAnyVoiceFew": "只需几秒录音,即可用任意声音朗读:可以是内置声音,也可以是媒体库中的样本。",
"speech.readsVoiceHeardClip": "用时间线上所选片段中的声音朗读",
"speech.readsVoiceHeardRecording": "用媒体库中某段录音里的声音朗读。模型会聆听从下方所选位置开始的前十秒。",
diff --git a/src/crates/concat/locales/zh-TW.json b/src/crates/concat/locales/zh-TW.json
index 1e2f0c77..6c82d3a8 100644
--- a/src/crates/concat/locales/zh-TW.json
+++ b/src/crates/concat/locales/zh-TW.json
@@ -17,6 +17,7 @@
"captions.oneTitlePerPhrase": "每個口說片語一個標題,依字詞本身的時間切分,放在第一條有空位的軌道上。",
"captions.oneTitlePerSentence": "每句一個標題,停留足夠閱讀的時間,依序排列在第一條有空位的軌道上。",
"captions.placement": "位置",
+ "captions.readingTheSound": "正在讀取片段的聲音",
"captions.transcribing": "正在轉錄",
"captions.typeWordsPutScreen": "輸入要顯示在螢幕上的文字…",
"captions.useVoiceFromTimeline": "使用時間軸中的語音",
@@ -91,6 +92,7 @@
"clipInspector.transitionInto": "進入 {0}",
"clipInspector.treatments": "處理方式",
"clipInspector.volume": "音量",
+ "common.aboutLeft": "約剩 {0}:{1}",
"common.adjust": "調整",
"common.apply": "套用",
"common.audio": "音訊",
@@ -813,6 +815,7 @@
"speech.importClipVoiceFirst": "請先匯入一個含有人聲的片段。",
"speech.italian": "義大利文",
"speech.japanese": "日文",
+ "speech.loadingTheModel": "正在載入語音模型",
"speech.long": "長",
"speech.longer": "更長",
"speech.male": "男聲",
@@ -826,6 +829,7 @@
"speech.pickSampleVoice": "選擇一段含有人聲的樣本",
"speech.portuguese": "葡萄牙語",
"speech.quick": "快速",
+ "speech.readingTheSample": "正在讀取聲音錄音",
"speech.readsAnyVoiceFew": "只需幾秒錄音,即可用任意聲音朗讀:可以是內建聲音,也可以是媒體庫中的樣本。",
"speech.readsVoiceHeardClip": "用時間軸上所選片段中的聲音朗讀",
"speech.readsVoiceHeardRecording": "用媒體庫中某段錄音裡的聲音朗讀。模型會聆聽從下方所選位置開始的前十秒。",
diff --git a/src/crates/concat/src/panes/captions.rs b/src/crates/concat/src/panes/captions.rs
index c3c8384d..f8198f97 100644
--- a/src/crates/concat/src/panes/captions.rs
+++ b/src/crates/concat/src/panes/captions.rs
@@ -17,11 +17,13 @@ use std::sync::Arc;
use concat_project::Command;
use concat_project::model::TextStyle;
+use concat_speech::JobStage;
use concat_speech::transcribe::Segment;
use crate::host::{on_ui_in_project, spawn_in_project};
use crate::i18n::{t, tf};
use crate::panes::Msg;
+use crate::panes::pace::{Pace, Stage, Timings};
use crate::panes::settings::installed;
use crate::studio::Studio;
use crate::ui::CaptionsSheetData;
@@ -55,22 +57,31 @@ pub enum CaptionsMsg {
/// Run the pass the sheet describes.
Begin,
Cancel,
- /// The transcriber's worker reporting where it is, in percent.
- Progress(i32),
+ /// The transcriber's worker reporting the stage it is in.
+ Progress(JobStage),
/// The transcriber's worker is done: what was said, or why not.
Finished(Result<Vec<Segment>, String>),
}
/// The clip a transcription is running over, as the finished segments
/// need it: where it starts on the timeline and how fast it plays.
-#[derive(Clone, Copy, Debug)]
+#[derive(Clone, Debug)]
struct Subject {
start: f64,
speed: f64,
/// The look chosen when the run began.
look: (f64, f64),
+ /// The transcriber model, and the thirty-second windows whisper reads
+ /// the sound in: what the run's timing is kept by.
+ model: String,
+ windows: f32,
}
+/// Seconds whisper takes over one thirty-second window, per megabyte of
+/// model, before a run on the model has been timed: about a second for
+/// base and ten for medium on eight cores of a recent CPU.
+const WINDOW_SECONDS_PER_MB: f32 = 1.0 / 160.0;
+
/// The captions sheet's state.
#[derive(Default)]
pub struct CaptionsPane {
@@ -91,11 +102,14 @@ pub struct CaptionsPane {
/// 0 small, 1 medium, 2 large.
pub size: usize,
pub running: bool,
- pub progress: f32,
+ /// How far the running transcription looks.
+ pub pace: Pace,
/// Why the last run failed, when it did.
pub message: String,
/// The clip under the running transcription.
subject: Option<Subject>,
+ /// How long each model took this session, for the next run's bar.
+ timings: Timings,
}
impl CaptionsPane {
@@ -149,12 +163,14 @@ impl CaptionsPane {
self.running = false;
self.open = false;
}
- CaptionsMsg::Progress(percent) => {
- self.progress = (percent as f32 / 100.0).clamp(0.0, 1.0);
- }
+ CaptionsMsg::Progress(stage) => self.pace.report(stage),
CaptionsMsg::Finished(result) => {
self.running = false;
let subject = self.subject.take();
+ if let (Ok(_), Some(subject)) = (&result, &subject) {
+ self.timings
+ .keep(&subject.model, self.pace.learned(subject.windows));
+ }
match result {
Ok(segments) => {
let Some(subject) = subject else {
@@ -216,6 +232,7 @@ impl CaptionsPane {
model,
placement: 0,
size: 1,
+ timings: std::mem::take(&mut self.timings),
..CaptionsPane::default()
};
}
@@ -263,9 +280,9 @@ impl CaptionsPane {
self.message = t("captions.clipHasNoFile");
return;
};
- let Some(model) = installed(&studio.settings.transcribers)
+ let Some((model, megabytes)) = installed(&studio.settings.transcribers)
.get(self.model)
- .map(|model| model.id.clone())
+ .map(|model| (model.id.clone(), model.megabytes))
else {
self.message = t("captions.downloadTranscriberModelFirst");
return;
@@ -275,24 +292,45 @@ impl CaptionsPane {
audio_stream: clip.audio_stream,
source_start: clip.source_start,
window: clip.duration * clip.speed,
- model_id: model,
+ model_id: model.clone(),
};
let dirs = studio.host.dirs.clone();
let transcriber = Arc::clone(&studio.host.transcriber);
+ // The run as the bar lays it out: the sound read in at a few
+ // hundred times its length, the model loaded unless it still is,
+ // and whisper a window at a time - at the pace the last run on
+ // this model measured, or the estimate by its size.
+ let windows = (request.window / 30.0).ceil().max(1.0) as f32;
+ let learned = self.timings.get(&model);
+ let loaded = studio.host.transcriber.loaded_model().as_deref() == Some(model.as_str());
+ let load = if loaded {
+ 0.0
+ } else {
+ learned.load.unwrap_or(0.5 + megabytes / 400.0)
+ };
+ let per_window = learned
+ .per_unit
+ .unwrap_or(0.3 + megabytes * WINDOW_SECONDS_PER_MB);
+ self.pace = Pace::new(vec![
+ (Stage::Reading, 0.2 + request.window as f32 / 400.0),
+ (Stage::Loading, load),
+ (Stage::Working, windows * per_window),
+ ]);
self.subject = Some(Subject {
start: clip.start,
speed: clip.speed,
look: self.look(),
+ model,
+ windows,
});
self.running = true;
- self.progress = 0.0;
self.message.clear();
let epoch = crate::host::project_epoch();
spawn_in_project(
move || {
- transcriber.transcribe(&dirs, &request, move |percent| {
+ transcriber.transcribe(&dirs, &request, move |stage| {
on_ui_in_project(epoch, move |studio, _, _| {
- studio.handle(Msg::Captions(CaptionsMsg::Progress(percent)));
+ studio.handle(Msg::Captions(CaptionsMsg::Progress(stage)));
});
})
},
@@ -326,7 +364,16 @@ impl CaptionsPane {
placement: self.placement as i32,
size: self.size as i32,
running: self.running,
- progress: self.progress,
+ progress: self.pace.floor,
+ ceiling: self.pace.ceiling,
+ step: self.pace.step,
+ left: self.pace.left,
+ stage: match self.pace.stage() {
+ Some(Stage::Loading) => t("captions.loadingTheModel"),
+ Some(Stage::Working) => t("captions.transcribing"),
+ _ => t("captions.readingTheSound"),
+ }
+ .into(),
ready: !installed(&studio.settings.transcribers).is_empty(),
message: self.message.as_str().into(),
}
diff --git a/src/crates/concat/src/panes/mod.rs b/src/crates/concat/src/panes/mod.rs
index caf71dd2..d9b9ded1 100644
--- a/src/crates/concat/src/panes/mod.rs
+++ b/src/crates/concat/src/panes/mod.rs
@@ -18,6 +18,7 @@ pub mod captions;
pub mod export;
pub mod media_bin;
pub mod monitor;
+pub mod pace;
pub mod project;
pub mod relink;
pub mod settings;
diff --git a/src/crates/concat/src/panes/pace.rs b/src/crates/concat/src/panes/pace.rs
new file mode 100644
index 00000000..920e3494
--- /dev/null
+++ b/src/crates/concat/src/panes/pace.rs
@@ -0,0 +1,310 @@
+// SPDX-License-Identifier: AGPL-3.0-or-later
+// SPDX-FileCopyrightText: 2026 Jareer and Concat contributors
+
+//! How far along a transcription or a read looks, from what it reports.
+//!
+//! The jobs report rarely: whisper once per thirty seconds of sound, Kokoro
+//! and Pocket once per sentence, and nothing at all while the sound is read
+//! in or the model loads. A bar fed only those numbers sits still for most
+//! of a run and then jumps. [`Pace`] lays the run out as its stages, each
+//! given room on the bar by how long it is expected to take, and for each
+//! report says three things: where the bar surely is (`floor`), where the
+//! step under way will leave it (`ceiling`), and how long that step should
+//! take (`step`). The bar moves from floor toward ceiling in that time and
+//! never reaches the ceiling before the job says so; see `Progress` in
+//! primitives/modal.slint.
+//!
+//! The expectations start as estimates from the model's size and the
+//! input's length, and are replaced by what the run measures once the work
+//! has reported once: from then on the step and the time left are timed,
+//! not guessed. [`Learned`] carries a finished run's timings to the next
+//! run on the same model.
+
+use std::collections::HashMap;
+use std::time::Instant;
+
+use concat_speech::JobStage;
+
+/// A stage of a run, as the plan names it.
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub enum Stage {
+ Reading,
+ Loading,
+ Working,
+}
+
+/// One run's progress, as the bar draws it.
+#[derive(Clone, Debug)]
+pub struct Pace {
+ /// The stages the run will report, in order, each with the seconds it
+ /// is expected to take.
+ plan: Vec<(Stage, f32)>,
+ /// The stage under way, by its row in `plan`.
+ at: Option<usize>,
+ /// When the stage under way began.
+ since: Instant,
+ /// When the work began, which is what its rate is timed from.
+ working: Option<Instant>,
+ /// How long the model took to load, when this run loaded it.
+ loaded_in: Option<f32>,
+ /// Where the bar surely is, `0..=1`.
+ pub floor: f32,
+ /// Where the step under way will leave it, `floor..=1`.
+ pub ceiling: f32,
+ /// Seconds the step under way is expected to take.
+ pub step: f32,
+ /// Seconds of the run still to go, once the work has been timed;
+ /// negative before.
+ pub left: f32,
+}
+
+impl Default for Pace {
+ fn default() -> Self {
+ Self::new(Vec::new())
+ }
+}
+
+impl Pace {
+ /// A run that will go through `plan`'s stages, in order. A stage the
+ /// run turns out to skip (a model already loaded) is passed over when
+ /// the next one is reported.
+ pub fn new(plan: Vec<(Stage, f32)>) -> Self {
+ let mut pace = Self {
+ plan,
+ at: None,
+ since: Instant::now(),
+ working: None,
+ loaded_in: None,
+ floor: 0.0,
+ ceiling: 0.0,
+ step: 0.0,
+ left: -1.0,
+ };
+ // Before the first report the run heads for the end of its first
+ // stage that takes any time.
+ if let Some(index) = pace.first() {
+ pace.ceiling = pace.span(index).1;
+ pace.step = pace.plan[index].1;
+ }
+ pace
+ }
+
+ /// The stage the run is in: the one it last reported, or before any
+ /// report the first that takes any time.
+ pub fn stage(&self) -> Option<Stage> {
+ self.at
+ .or_else(|| self.first())
+ .map(|index| self.plan[index].0)
+ }
+
+ /// The first stage in the plan expected to take any time.
+ fn first(&self) -> Option<usize> {
+ self.plan.iter().position(|&(_, seconds)| seconds > 0.0)
+ }
+
+ /// Where on the bar stage `index` runs, from and to.
+ fn span(&self, index: usize) -> (f32, f32) {
+ let total: f32 = self.plan.iter().map(|(_, seconds)| seconds).sum();
+ if total <= 0.0 {
+ return (0.0, 1.0);
+ }
+ let before: f32 = self.plan[..index].iter().map(|(_, seconds)| seconds).sum();
+ let start = before / total;
+ (start, start + self.plan[index].1 / total)
+ }
+
+ /// Takes in one report from the run.
+ pub fn report(&mut self, stage: JobStage) {
+ let named = match stage {
+ JobStage::Reading => Stage::Reading,
+ JobStage::Loading => Stage::Loading,
+ JobStage::Working { .. } => Stage::Working,
+ };
+ let Some(index) = self.plan.iter().position(|(planned, _)| *planned == named) else {
+ return;
+ };
+ if self.at != Some(index) {
+ if self.at.is_some_and(|at| self.plan[at].0 == Stage::Loading) {
+ self.loaded_in = Some(self.since.elapsed().as_secs_f32());
+ }
+ self.at = Some(index);
+ self.since = Instant::now();
+ }
+ let (start, end) = self.span(index);
+ let (done, next) = match stage {
+ JobStage::Working { done, next } => {
+ let done = done.clamp(0.0, 1.0);
+ (done, next.clamp(done, 1.0))
+ }
+ _ => (0.0, 1.0),
+ };
+ self.floor = self.floor.max(start + done * (end - start));
+ self.ceiling = (start + next * (end - start)).max(self.floor);
+ let planned = self.plan[index].1;
+ self.step = planned * (next - done);
+ self.left = -1.0;
+ if named == Stage::Working {
+ let working = *self.working.get_or_insert_with(Instant::now);
+ // Timed once a little of the work is done: before that the
+ // rate is mostly the time to the first report.
+ if done >= 0.02 {
+ let per = working.elapsed().as_secs_f32() / done;
+ self.step = per * (next - done);
+ self.left = per * (1.0 - done);
+ }
+ }
+ }
+
+ /// What this run measured, for the next one on the same model: the
+ /// load, when it loaded, and the work's seconds per `units` of input.
+ pub fn learned(&self, units: f32) -> Learned {
+ Learned {
+ load: self.loaded_in,
+ per_unit: self
+ .working
+ .filter(|_| units > 0.0)
+ .map(|working| working.elapsed().as_secs_f32() / units),
+ }
+ }
+}
+
+/// A finished run's timings, kept for the next run on the same model.
+#[derive(Clone, Copy, Debug, Default, PartialEq)]
+pub struct Learned {
+ /// Seconds the model took to load.
+ pub load: Option<f32>,
+ /// Seconds of work per unit of input: per thirty-second window for a
+ /// transcription, per character for a read.
+ pub per_unit: Option<f32>,
+}
+
+/// The timings learned this session, by model id.
+#[derive(Clone, Debug, Default)]
+pub struct Timings(HashMap<String, Learned>);
+
+impl Timings {
+ pub fn get(&self, model: &str) -> Learned {
+ self.0.get(model).copied().unwrap_or_default()
+ }
+
+ /// Keeps what a finished run measured, over what an earlier one did; a
+ /// run that did not load the model keeps the earlier load.
+ pub fn keep(&mut self, model: &str, learned: Learned) {
+ let kept = self.0.entry(model.to_owned()).or_default();
+ kept.load = learned.load.or(kept.load);
+ kept.per_unit = learned.per_unit.or(kept.per_unit);
+ }
+}
+
+#[cfg(test)]
+mod tests {
+ use super::*;
+
+ fn close(a: f32, b: f32) -> bool {
+ (a - b).abs() < 1e-4
+ }
+
+ /// Each stage has room on the bar by its expected time, and a stage's
+ /// report heads the bar for the end of it.
+ #[test]
+ fn stages_take_room_by_their_expected_time() {
+ let mut pace = Pace::new(vec![
+ (Stage::Reading, 1.0),
+ (Stage::Loading, 3.0),
+ (Stage::Working, 6.0),
+ ]);
+ assert!(close(pace.ceiling, 0.1));
+ pace.report(JobStage::Reading);
+ assert!(close(pace.floor, 0.0) && close(pace.ceiling, 0.1));
+ assert!(close(pace.step, 1.0));
+ pace.report(JobStage::Loading);
+ assert!(close(pace.floor, 0.1) && close(pace.ceiling, 0.4));
+ assert!(close(pace.step, 3.0));
+ pace.report(JobStage::Working {
+ done: 0.0,
+ next: 0.5,
+ });
+ assert!(close(pace.floor, 0.4) && close(pace.ceiling, 0.7));
+ assert!(close(pace.step, 3.0));
+ assert!(pace.left < 0.0, "nothing timed yet");
+ pace.report(JobStage::Working {
+ done: 0.5,
+ next: 1.0,
+ });
+ assert!(close(pace.floor, 0.7) && close(pace.ceiling, 1.0));
+ assert!(pace.left >= 0.0, "timed once the work has reported");
+ }
+
+ /// A stage the run skips - the model was still loaded - is passed
+ /// over, and the bar never goes back.
+ #[test]
+ fn a_skipped_stage_is_passed_over_and_the_bar_never_goes_back() {
+ let mut pace = Pace::new(vec![
+ (Stage::Reading, 1.0),
+ (Stage::Loading, 1.0),
+ (Stage::Working, 2.0),
+ ]);
+ pace.report(JobStage::Reading);
+ pace.report(JobStage::Working {
+ done: 0.0,
+ next: 1.0,
+ });
+ assert!(close(pace.floor, 0.5) && close(pace.ceiling, 1.0));
+ pace.report(JobStage::Reading);
+ assert!(close(pace.floor, 0.5), "a late report cannot move it back");
+ assert!(pace.ceiling >= pace.floor);
+ }
+
+ /// A model still loaded takes no time: before any report the run is
+ /// in, and heads for the end of, the first stage that does.
+ #[test]
+ fn a_stage_of_no_time_is_not_the_one_shown_first() {
+ let pace = Pace::new(vec![(Stage::Loading, 0.0), (Stage::Working, 4.0)]);
+ assert_eq!(pace.stage(), Some(Stage::Working));
+ assert!(close(pace.ceiling, 1.0) && close(pace.step, 4.0));
+ }
+
+ /// Reports outside the plan, or with numbers out of range, change
+ /// nothing they should not.
+ #[test]
+ fn reports_out_of_range_are_held_in() {
+ let mut pace = Pace::new(vec![(Stage::Working, 1.0)]);
+ pace.report(JobStage::Loading);
+ assert!(close(pace.floor, 0.0));
+ pace.report(JobStage::Working {
+ done: 1.5,
+ next: 0.2,
+ });
+ assert!(close(pace.floor, 1.0) && close(pace.ceiling, 1.0));
+ let empty = Pace::default();
+ assert!(close(empty.floor, 0.0) && close(empty.ceiling, 0.0));
+ }
+
+ /// A run that did not load keeps the earlier run's load.
+ #[test]
+ fn timings_keep_the_last_measure_of_each() {
+ let mut timings = Timings::default();
+ timings.keep(
+ "base",
+ Learned {
+ load: Some(2.0),
+ per_unit: Some(1.0),
+ },
+ );
+ timings.keep(
+ "base",
+ Learned {
+ load: None,
+ per_unit: Some(0.5),
+ },
+ );
+ assert_eq!(
+ timings.get("base"),
+ Learned {
+ load: Some(2.0),
+ per_unit: Some(0.5),
+ }
+ );
+ assert_eq!(timings.get("other"), Learned::default());
+ }
+}
diff --git a/src/crates/concat/src/panes/speech.rs b/src/crates/concat/src/panes/speech.rs
index 9cbcaf58..bf5a2ff5 100644
--- a/src/crates/concat/src/panes/speech.rs
+++ b/src/crates/concat/src/panes/speech.rs
@@ -25,6 +25,7 @@ use std::sync::Arc;
use concat_host::media::{self, MediaSummary};
use concat_project::Command;
use concat_project::model::{ClipKind, MediaItem, MediaOrigin};
+use concat_speech::JobStage;
use concat_speech::tts::{
CHATTERBOX_CLONE, Family, POCKET_CLONE, Reference, VoiceInfo, family_of, is_clone,
};
@@ -35,6 +36,7 @@ use crate::host::{on_ui_in_project, spawn_in_project};
use crate::i18n::{t, tf};
use crate::panes::Msg;
use crate::panes::captions::CHARS_PER_SECOND;
+use crate::panes::pace::{Pace, Stage, Timings};
use crate::panes::settings::installed;
use crate::studio::Studio;
use crate::ui::SpeechSheetData;
@@ -54,6 +56,16 @@ const STEPS: [i32; 3] = [3, 5, 8];
/// The breath between sentences the sheet opens on: sherpa's own.
const DEFAULT_PAUSES: f32 = 0.2;
+/// Seconds a model takes to read one second of speech, before a read with
+/// it has been timed: about what each took on a recent CPU.
+fn reading_pace(family: Family) -> f32 {
+ match family {
+ Family::Kokoro => 0.35,
+ Family::Pocket => 0.3,
+ Family::Chatterbox => 1.6,
+ }
+}
+
/// Everything that can happen to the speech sheet.
#[derive(Clone, Debug)]
pub enum SpeechMsg {
@@ -79,7 +91,7 @@ pub enum SpeechMsg {
Begin,
Cancel,
/// The synthesiser's worker reporting where it is, `0..=1`.
- Progress(f32),
+ Progress(JobStage),
/// The synthesiser's worker is done: the file it wrote, probed, or
/// why not.
Finished(Box<Result<MediaSummary, String>>),
@@ -116,7 +128,13 @@ pub struct SpeechPane {
/// a recording is not always where the speaking is.
pub reference_start: f64,
pub running: bool,
- pub progress: f32,
+ /// How far the running read looks.
+ pub pace: Pace,
+ /// The model and the length of the running read: what its timing is
+ /// kept by.
+ reading: Option<(String, f32)>,
+ /// How long each model took this session, for the next read's bar.
+ timings: Timings,
pub message: String,
/// Where the running read will land on the timeline, in seconds.
landing: f64,
@@ -157,6 +175,7 @@ impl SpeechPane {
sample: 0,
reference_start: 0.0,
speakers: std::mem::take(&mut self.speakers),
+ timings: std::mem::take(&mut self.timings),
..SpeechPane::default()
};
// The voice chosen last time, if the model reads with it;
@@ -215,9 +234,12 @@ impl SpeechPane {
self.running = false;
self.open = false;
}
- SpeechMsg::Progress(fraction) => self.progress = fraction.clamp(0.0, 1.0),
+ SpeechMsg::Progress(stage) => self.pace.report(stage),
SpeechMsg::Finished(result) => {
self.running = false;
+ if let (Ok(_), Some((model, chars))) = (result.as_ref(), self.reading.take()) {
+ self.timings.keep(&model, self.pace.learned(chars));
+ }
match *result {
Ok(summary) => {
// Marked as read aloud: the bin shelves it under
@@ -258,9 +280,9 @@ impl SpeechPane {
self.message = t("speech.nothingToReadYet");
return;
}
- let Some(model) = installed(&studio.settings.voices)
+ let Some((model, megabytes)) = installed(&studio.settings.voices)
.get(self.model)
- .map(|model| model.id.clone())
+ .map(|model| (model.id.clone(), model.megabytes))
else {
self.message = t("speech.downloadVoiceModelFirst");
return;
@@ -320,6 +342,28 @@ impl SpeechPane {
.and_then(|id| studio.clip(id))
.map(|clip| clip.start)
.unwrap_or(f64::from(studio.playhead));
+ // The read as the bar lays it out: the model loaded unless it
+ // still is, the recording the voice comes from read in (Kokoro
+ // has none), and the reading at the pace the last read with this
+ // model measured, or the model's usual pace.
+ let chars = text.chars().count() as f32;
+ let learned = self.timings.get(&model);
+ let loaded = studio.host.speech.loaded_model().as_deref() == Some(model.as_str());
+ let load = if loaded {
+ 0.0
+ } else {
+ learned.load.unwrap_or(0.5 + megabytes / 150.0)
+ };
+ let per_char = learned
+ .per_unit
+ .unwrap_or(reading_pace(self.family) / CHARS_PER_SECOND / self.rate());
+ let mut plan = vec![(Stage::Loading, load)];
+ if self.family != Family::Kokoro {
+ plan.push((Stage::Reading, 0.3));
+ }
+ plan.push((Stage::Working, chars * per_char));
+ self.pace = Pace::new(plan);
+ self.reading = Some((model.clone(), chars));
let request = concat_speech::tts::SpeakRequest {
model_id: model,
voice,
@@ -334,14 +378,13 @@ impl SpeechPane {
let dirs = studio.host.dirs.clone();
let speech = Arc::clone(&studio.host.speech);
self.running = true;
- self.progress = 0.0;
self.message.clear();
let epoch = crate::host::project_epoch();
spawn_in_project(
move || {
- let spoken = speech.speak(&dirs, &request, move |fraction| {
+ let spoken = speech.speak(&dirs, &request, move |stage| {
on_ui_in_project(epoch, move |studio, _, _| {
- studio.handle(Msg::Speech(SpeechMsg::Progress(fraction)));
+ studio.handle(Msg::Speech(SpeechMsg::Progress(stage)));
});
})?;
let summary = media::probe(&spoken.path)?;
@@ -560,7 +603,16 @@ impl SpeechPane {
reference_length: length as f32,
reference_start: self.reference_offset(length) as f32,
running: self.running,
- progress: self.progress,
+ progress: self.pace.floor,
+ ceiling: self.pace.ceiling,
+ step: self.pace.step,
+ left: self.pace.left,
+ stage: match self.pace.stage() {
+ Some(Stage::Reading) => t("speech.readingTheSample"),
+ Some(Stage::Working) => t("speech.generatingVoice"),
+ _ => t("speech.loadingTheModel"),
+ }
+ .into(),
ready: !installed(&studio.settings.voices).is_empty(),
estimate: self.estimate(studio).into(),
message: self.message.as_str().into(),
diff --git a/src/crates/concat/src/studio.rs b/src/crates/concat/src/studio.rs
index 593d8478..00479419 100644
--- a/src/crates/concat/src/studio.rs
+++ b/src/crates/concat/src/studio.rs
@@ -7247,9 +7247,9 @@ impl Studio {
self.art_stamp = None;
self.proxied.clear();
self.captions.running = false;
- self.captions.progress = 0.0;
+ self.captions.pace = Default::default();
self.speech.running = false;
- self.speech.progress = 0.0;
+ self.speech.pace = Default::default();
self.source = None;
// A take running as the project closes stops, keeping its file;
// the next project's speakers are not left silent.
diff --git a/src/crates/concat/ui/dialogs/captions.slint b/src/crates/concat/ui/dialogs/captions.slint
index a48dcde2..c23b6013 100644
--- a/src/crates/concat/ui/dialogs/captions.slint
+++ b/src/crates/concat/ui/dialogs/captions.slint
@@ -40,8 +40,17 @@ export struct CaptionsSheetData {
/// 0 Small, 1 Medium, 2 Large
size: int,
running: bool,
- /// 0..1
+ /// 0..1: where the transcription surely is.
progress: float,
+ /// Where the step under way leaves the bar, 0..1.
+ ceiling: float,
+ /// Seconds that step should take.
+ step: float,
+ /// Seconds of the run still to go, once the work has been timed;
+ /// negative before.
+ left: float,
+ /// What the run is doing, in words.
+ stage: string,
/// A transcriber model is installed.
ready: bool,
/// Why the last run failed, when it did.
@@ -265,28 +274,48 @@ export component CaptionsDialog inherits Modal {
}
}
- if root.data.running: HorizontalLayout {
+ if root.data.running: VerticalLayout {
spacing: 8px;
- Text {
- horizontal-stretch: 1;
- text: root.data.progress < 0.02 ? I18n.t("captions.loadingTheModel") : I18n.t("captions.transcribing");
- color: Theme.fg;
- font-size: Theme.fs;
- vertical-alignment: center;
+ row := HorizontalLayout {
+ spacing: 8px;
+ // The time left, once the work has been timed: counted down
+ // from the last report, and gone in the last second.
+ property <int> left: Math.ceil(Math.max(0, root.data.left - bar.since / 1s));
+
+ Text {
+ horizontal-stretch: 1;
+ text: root.data.stage;
+ color: Theme.fg;
+ font-size: Theme.fs;
+ vertical-alignment: center;
+ }
+
+ if root.data.left >= 0 && row.left > 0: Text {
+ text: I18n.t2("common.aboutLeft", "\{Math.floor(row.left / 60)}",
+ (Math.mod(row.left, 60) < 10 ? "0" : "") + "\{Math.mod(row.left, 60)}");
+ color: Theme.fg-dim;
+ font-size: Theme.fs;
+ vertical-alignment: center;
+ }
+
+ Text {
+ text: "\{Math.floor(bar.shown * 100)}%";
+ color: Theme.fg-muted;
+ font-family: Theme.font-technical;
+ font-size: Theme.fs;
+ vertical-alignment: center;
+ }
}
- Text {
- text: "\{Math.round(root.data.progress * 100)}%";
- color: Theme.fg-muted;
- font-family: Theme.font-technical;
- font-size: Theme.fs;
- vertical-alignment: center;
+ bar := Progress {
+ value: root.data.progress;
+ ceiling: root.data.ceiling;
+ step: root.data.step * 1s;
+ live: true;
}
}
- if root.data.running: Progress { value: root.data.progress; }
-
if root.data.running: HorizontalLayout {
alignment: start;
Button {
diff --git a/src/crates/concat/ui/dialogs/export.slint b/src/crates/concat/ui/dialogs/export.slint
index 6be2eb27..0d2fe92a 100644
--- a/src/crates/concat/ui/dialogs/export.slint
+++ b/src/crates/concat/ui/dialogs/export.slint
@@ -610,7 +610,7 @@ component ExportFooter inherits VerticalLayout {
}
}
- if root.running: Progress { value: root.data.progress; }
+ if root.running: Progress { value: root.data.progress; live: true; }
if root.running: HorizontalLayout {
alignment: start;
diff --git a/src/crates/concat/ui/dialogs/speech.slint b/src/crates/concat/ui/dialogs/speech.slint
index ae2c6a60..53d8d37e 100644
--- a/src/crates/concat/ui/dialogs/speech.slint
+++ b/src/crates/concat/ui/dialogs/speech.slint
@@ -56,8 +56,17 @@ export struct SpeechSheetData {
/// Seconds into that sample the speaking starts.
reference-start: float,
running: bool,
- /// 0..1, as sentences complete.
+ /// 0..1: where the read surely is.
progress: float,
+ /// Where the step under way leaves the bar, 0..1.
+ ceiling: float,
+ /// Seconds that step should take.
+ step: float,
+ /// Seconds of the run still to go, once the work has been timed;
+ /// negative before.
+ left: float,
+ /// What the run is doing, in words.
+ stage: string,
/// A voice model is installed.
ready: bool,
/// "About 0:12 in Heart", or what stops it being read.
@@ -589,38 +598,58 @@ export component SpeechDialog inherits Modal {
}
}
- if root.running: HorizontalLayout {
+ if root.running: VerticalLayout {
spacing: 8px;
- Text {
- horizontal-stretch: 1;
- text: I18n.t("speech.generatingVoice");
- color: Theme.fg;
- font-size: Theme.fs;
- vertical-alignment: center;
- }
+ row := HorizontalLayout {
+ spacing: 8px;
+ // The time left, once the work has been timed: counted down
+ // from the last report, and gone in the last second.
+ property <int> left: Math.ceil(Math.max(0, root.data.left - bar.since / 1s));
- Text {
- text: "\{Math.round(root.data.progress * 100)}%";
- color: Theme.fg-muted;
- font-family: Theme.font-technical;
- font-size: Theme.fs;
- vertical-alignment: center;
+ Text {
+ horizontal-stretch: 1;
+ text: root.data.stage;
+ color: Theme.fg;
+ font-size: Theme.fs;
+ vertical-alignment: center;
+ }
+
+ if root.data.left >= 0 && row.left > 0: Text {
+ text: I18n.t2("common.aboutLeft", "\{Math.floor(row.left / 60)}",
+ (Math.mod(row.left, 60) < 10 ? "0" : "") + "\{Math.mod(row.left, 60)}");
+ color: Theme.fg-dim;
+ font-size: Theme.fs;
+ vertical-alignment: center;
+ }
+
+ Text {
+ text: "\{Math.floor(bar.shown * 100)}%";
+ color: Theme.fg-muted;
+ font-family: Theme.font-technical;
+ font-size: Theme.fs;
+ vertical-alignment: center;
+ }
+
+ Text {
+ text: "\{Math.floor(root.elapsed / 60)}:"
+ + (Math.mod(root.elapsed, 60) < 10 ? "0" : "")
+ + "\{Math.mod(root.elapsed, 60)}";
+ color: Theme.fg-dim;
+ font-family: Theme.font-technical;
+ font-size: Theme.fs;
+ vertical-alignment: center;
+ }
}
- Text {
- text: "\{Math.floor(root.elapsed / 60)}:"
- + (Math.mod(root.elapsed, 60) < 10 ? "0" : "")
- + "\{Math.mod(root.elapsed, 60)}";
- color: Theme.fg-dim;
- font-family: Theme.font-technical;
- font-size: Theme.fs;
- vertical-alignment: center;
+ bar := Progress {
+ value: root.data.progress;
+ ceiling: root.data.ceiling;
+ step: root.data.step * 1s;
+ live: true;
}
}
- if root.running: Progress { value: root.data.progress; }
-
if root.running: HorizontalLayout {
alignment: start;
Button {
diff --git a/src/crates/concat/ui/primitives/modal.slint b/src/crates/concat/ui/primitives/modal.slint
index cad307e8..1d303b19 100644
--- a/src/crates/concat/ui/primitives/modal.slint
+++ b/src/crates/concat/ui/primitives/modal.slint
@@ -306,14 +306,52 @@ export component CopyChip inherits Rectangle {
// A determinate bar. Nothing indeterminate: every job in this app can say how
// far along it is, and a barber's pole for one that can is a lie.
+//
+// What a job says can be coarse - a transcription reports once per thirty
+// seconds of sound, a read once per sentence - so a job that knows where
+// the step under way ends gives it as `ceiling`, and the time the step
+// should take as `step`. The fill moves from `value` toward `ceiling` over
+// that time, steadily for nine tenths of the way and then ever slower, so
+// it never reaches the ceiling before the job reports it has. A report
+// ahead of the fill slides it forward rather than jumping. While `live`, a
+// sheen runs along the fill to say the job is still at work, which a fill
+// waiting on a report would not; the sheen never moves the fill.
export component Progress inherits Rectangle {
in property <float> value;
+ /// Where the step under way leaves the bar; `value` when the job does
+ /// not say.
+ in property <float> ceiling: root.value;
+ /// How long that step should take.
+ in property <duration> step;
/// Finished: the bar turns green rather than staying accent, so a glance
/// tells you it is over without reading the row above it.
in property <bool> done;
+ /// The job is running: the sheen runs along the fill.
+ in property <bool> live;
+ /// Where the fill is, 0..1: what a percentage beside the bar reads.
+ out property <float> shown: Math.clamp(
+ root.ceiling > root.value && root.step > 0ms
+ ? root.value + (root.ceiling - root.value) * root.creep((animation-tick() - root.mark) / root.step)
+ : root.value,
+ 0, 1);
+ /// The time since the last report, for a countdown beside the bar.
+ out property <duration> since: animation-tick() - root.mark;
+
+ /// When the job last reported.
+ property <duration> mark;
+ init => { root.mark = animation-tick(); }
+ changed value => { root.mark = animation-tick(); }
+ changed ceiling => { root.mark = animation-tick(); }
+
+ /// How far toward the ceiling the fill is `u` steps in: in a straight
+ /// line to nine tenths at one step, then half the rest by two, and so
+ /// on, never the whole.
+ pure function creep(u: float) -> float {
+ return u <= 1 ? 0.9 * Math.max(u, 0) : 0.9 + 0.1 * (u - 1) / u;
+ }
- height: 4px;
- border-radius: 2px;
+ height: 8px;
+ border-radius: self.height / 2;
background: Theme.well;
border-width: 1px;
border-color: Theme.line;
@@ -322,10 +360,21 @@ export component Progress inherits Rectangle {
Rectangle {
x: 0;
y: 0;
- width: parent.width * Math.clamp(root.value, 0, 1);
+ // Never thinner than it is tall once it has begun, so its end is
+ // round from the first step.
+ width: root.shown > 0 ? Math.max(parent.height, parent.width * root.shown) : 0px;
+ animate width { duration: 280ms; easing: ease-out; }
height: parent.height;
- border-radius: 2px;
+ border-radius: self.height / 2;
background: root.done ? Theme.green : Theme.accent;
+ clip: true;
+
+ if root.live && !root.done: Rectangle {
+ property <float> phase: Math.mod(animation-tick() / 1ms, 1800) / 1800;
+ width: Math.max(48px, root.width / 4);
+ x: (parent.width + self.width) * self.phase - self.width;
+ background: @linear-gradient(90deg, #ffffff00 0%, #ffffff59 50%, #ffffff00 100%);
+ }
}
}