different result between stable whisper and whisper

[Atefeh197]

why some times the transcription by whisper and stable whisper are different?

How I can get the same result?
I think the whisper does better, does s whisper_stable use the whisper model?

[jianfch]

Hello,
what were lines you ran to for both?

[Atefeh197]

Hi
Thank you,

model_stable = stable_whisper.load_model('medium')
result = model_stable.transcribe(audio_file)
print(result.text)

model_whisper = whisper.load_model('medium')
result = model_whisper.transcribe(audio_file)
print(result["text"])

[jianfch]

If your Stable-ts version is not 2.7+, try:

result = model_whisper.transcribe(audio_file, suppress_ts_tokens=False)

You can also try:

result = model_whisper.transcribe(audio_file, mel_first=True)