word_timestamps=True

App/Transcription/Utils/audio_transcription.py

@@ -15,7 +15,9 @@ def transcribe_file(state, file_path, model_size="tiny"):
         meta=metadata.dict(),
     )
     model = WhisperModel(model_size, device="cpu", compute_type="int8")
-    segments, info = model.transcribe(file_path, beam_size=5, vad_filter=True)
+    segments, info = model.transcribe(
+        file_path, beam_size=5, vad_filter=True, word_timestamps=True
+    )
 
     total_duration = round(info.duration, 2)
     metadata.logs = (
@@ -39,7 +41,6 @@ def transcribe_file(state, file_path, model_size="tiny"):
                         "end": word.end,
                         "text": word.text,
                     }
-                    # time_stamp = "[%.2fs -> %.2fs]" % (segment.start, segment.end)
                     result.append(temp)
             metadata.logs = "Transcribing.."
             metadata.percentage = f"{((segment.end / total_duration)*100)}"