Merge pull request huggingface#31 from xenova/whisper-timestamps

Add support for timestamped speech-to-text (w/ whisper)
ocavue · Mar 19, 2023 · f10106a · f10106a
2 parents 4ad2581 + ba9ee82
commit f10106a
Show file tree

Hide file tree

Showing 12 changed files with 1,033 additions and 184 deletions.
diff --git a/assets/js/worker.js b/assets/js/worker.js
@@ -310,6 +310,10 @@ async function speech_to_text(data) {
  })
 
  return await pipeline(data.audio, {
+ // Choose good defaults for the demo
+ chunk_length_s: 30,
+ stride_length_s: 5,
+
  ...data.generation,
  callback_function: function (beams) {
  const decodedText = pipeline.tokenizer.decode(beams[0].output_token_ids, {
@@ -322,10 +326,6 @@ async function speech_to_text(data) {
  data: decodedText.trim()
  });
  }
- }, {
- // Choose good defaults for the demo
- chunk_length_s: 30,
- stride_length_s: 5,
  })
 }