Using English Models = "ValueError: This model doesn't have language tokens so it can't perform lang id"

[ThioJoe]

I'm using the shorter code example from the readme - the first one under "Python usage".

It works except if I try to set the model to any of the ".en" (English only) ones, in which case I get the error:

ValueError: This model doesn't have language tokens so it can't perform lang id

Using the english specific model does work when using whisper via command line. Does something more need to be done if running it in a script?

Here's the full traceback:

Traceback (most recent call last):
  File "D:\whateverpath\whisper\myscript.py", line 10, in <module>
    result = model.transcribe("audio.mp3")
  File "D:\whateverpath\whisper\transcribe.py", line 82, in transcribe
    _, probs = model.detect_language(segment)
  File "C:\Users\Joe\miniconda3\envs\whisper\lib\site-packages\torch\autograd\grad_mode.py", line 27, in decorate_context
    return func(*args, **kwargs)
  File "D:\whateverpath\whisper\decoding.py", line 35, in detect_language
    raise ValueError(f"This model doesn't have language tokens so it can't perform lang id")
ValueError: This model doesn't have language tokens so it can't perform lang id

[astrowonk]

Try model.transcribe("file.wav",language='english')

Using small.en only slightly faster than small on my M1 Mac Mini, and actually slightly less accurate on my one self-recorded example audio file.

[ThioJoe]

Yep that worked

[jayavanth]

If you're using model.decode, pass it to DecodingOptions

options = whisper.DecodingOptions(language='english')

[OrbitPeppermint]

We had the same problem, fixed it by updating whisper to the latest Version with
pip install -U openai-whisper