Improved whisper params

nomadkaraoke · Jul 10, 2024 · a15b68d · a15b68d
1 parent 6a273ee
commit a15b68d
Show file tree

Hide file tree

Showing 3 changed files with 468 additions and 453 deletions.
diff --git a/lyrics_transcriber/transcriber.py b/lyrics_transcriber/transcriber.py
@@ -786,7 +786,7 @@ def transcribe(self):
         self.logger.debug(f"no cached transcription file found, running whisper transcribe with model: {self.transcription_model}")
         audio = whisper.load_audio(self.audio_filepath)
         model = whisper.load_model(self.transcription_model, device="cpu")
-        result = whisper.transcribe(model, audio, language="en")
+        result = whisper.transcribe(model, audio, language="en", vad="auditok", beam_size=5, temperature=0.2, best_of=5)
 
         self.logger.debug(f"transcription complete, performing post-processing cleanup")