How can I properly use whisper_timestamped in Python3

Question

How can I use whisper-timestamped's transcribe method? I have audio file "example.mp3", I want to generate it's timestamped transcription, as I have done my research open Ai's whisper-timestamped is great option. when i try to use transcribe method I get error.

my code:

import whisper_timestamped as whisper
import json


audio = whisper.load_audio('example.mp3')

model = whisper.load_model("small", device="cpu")

result = whisper.transcribe(model, audio, language="en")

print(json.dumps(result, indent = 2, ensure_ascii = False))

error:

  File "/home/luka/PycharmProjects/CreateAutoShorts/.venv/lib/python3.12/site-packages/whisper_timestamped-1.15.4-py3.12.egg/whisper_timestamped/transcribe.py", line 882, in <lambda>
    lambda layer, ins, outs, index=j: hook_attention_weights(layer, ins, outs, index))
                                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/luka/PycharmProjects/CreateAutoShorts/.venv/lib/python3.12/site-packages/whisper_timestamped-1.15.4-py3.12.egg/whisper_timestamped/transcribe.py", line 777, in hook_attention_weights
    if w.shape[-2] > 1:
       ^^^^^^^
AttributeError: 'NoneType' object has no attribute 'shape'

I tried reinstalling packages but without any success

Answer 1

I faced the same issue, using old version(20231117) of openai-whisper worked for me.