Openvino whisper transcription fails

hello!
I open a second topic because this issue is on another computer and is different from the issue in my other topic.
my computer is a macbook 2012 with w10 (in english)
I tried first with audacity 3.5.0 and the transcription failed.
Then i installed audacity 3.5.1 and the fix for openvino whisper to work with 3.5.1, and same fail happens.
here are the logs in audacity:

16:07:01: Audacity 3.5.1
16:07:01: Error: Failed to load shared library 'avformat-60.dll' (error 126: The specified module could not be found.)
16:07:01: Error: Failed to load shared library 'avformat-59.dll' (error 126: The specified module could not be found.)
16:07:01: Error: Failed to load shared library 'avformat-58.dll' (error 126: The specified module could not be found.)
16:07:01: Error: Failed to load shared library 'avformat-57.dll' (error 126: The specified module could not be found.)
16:07:01: Error: Failed to load shared library 'avformat-55.dll' (error 126: The specified module could not be found.)
16:07:06: Error: Failed to load shared library 'avformat-60.dll' (error 126: The specified module could not be found.)
16:07:06: Error: Failed to load shared library 'avformat-59.dll' (error 126: The specified module could not be found.)
16:07:06: Error: Failed to load shared library 'avformat-58.dll' (error 126: The specified module could not be found.)
16:07:06: Error: Failed to load shared library 'avformat-57.dll' (error 126: The specified module could not be found.)
16:07:06: Error: Failed to load shared library 'avformat-55.dll' (error 126: The specified module could not be found.)
16:07:15: File name is C:\Users\Administrator\Downloads\YogaDoux_FB2_07mn_REC.wav
16:07:15: Mime type is *
16:07:15: Opening with libsndfile
16:07:15: Open(C:\Users\Administrator\Downloads\YogaDoux_FB2_07mn_REC.wav) succeeded
16:07:16: Operation 'Importing WAV (Microsoft)' took 1.281000 seconds. Poll was called 133 times and took 0.029393 seconds. Yield was called 15 times and took 0.009255 seconds.
16:07:16: Operation 'Music Information Retrieval' took 0.000000 seconds. Poll was called 0 times and took 0.000000 seconds. Yield was called 0 times and took 0.000000 seconds.
16:07:24: Operation 'Pre-processing' took 0.000000 seconds. Poll was called 0 times and took 0.000000 seconds. Yield was called 0 times and took 0.000000 seconds.
16:07:28: Error: In Whisper Transcription Effect, exception: whisper.cpp context creation / initialization failed
16:07:35: Operation 'OpenVINO Whisper Transcription' took 3.821000 seconds. Poll was called 1082 times and took 0.143049 seconds. Yield was called 42 times and took 0.060357 seconds.
16:07:28: Error: whisper_init_from_file_with_params(C:\Program Files\Audacity\openvino-models\ggml-base.bin, ...) failed

can someone please help me?
thanks!

Hi @batmad,

Does this happen when ‘CPU’ is selected in the transcription settings, or do you have it set to another device?

Thanks,
Ryan

the only choice i have is CPU in this list!

Got it. Unfortunately, I’ll need to have you run Audacity from cmd.exe (command prompt) and save a log in order to get more information about the specific error from whisper.cpp.

Can you open cmd.exe (command prompt), and do the following?

cd C:\Program Files\Audacity\

Audacity.exe > %appdata%\audacity\audacity_log_from_cmd.txt

Once Audacity comes up, run the transcription effect as usual, triggering the error. You can close Audacity once the error has been triggered.

Go to %appdata%\audacity\ (copy and paste that into Windows explorer), and you should find the audacity_log_from_cmd.txt that was saved.

Can you upload this here?

Thanks!

here is the log:

is_whisper_model_present(tiny)
is_whisper_model_present: returning false because ggml-tiny.bin doesn't exist.
is_whisper_model_present(tiny.en)
is_whisper_model_present: returning false because ggml-tiny.en.bin doesn't exist.
is_whisper_model_present(base)
is_whisper_model_present: returning true
is_whisper_model_present(base.en)
is_whisper_model_present: returning false because ggml-base.en.bin doesn't exist.
is_whisper_model_present(small)
is_whisper_model_present: returning false because ggml-small.bin doesn't exist.
is_whisper_model_present(small.en)
is_whisper_model_present: returning false because ggml-small.en.bin doesn't exist.
is_whisper_model_present(small.en-tdrz)
is_whisper_model_present: returning false because ggml-small.en-tdrz.bin doesn't exist.
is_whisper_model_present(medium)
is_whisper_model_present: returning false because ggml-medium.bin doesn't exist.
is_whisper_model_present(medium.en)
is_whisper_model_present: returning false because ggml-medium.en.bin doesn't exist.
is_whisper_model_present(large-v1)
is_whisper_model_present: returning false because ggml-large-v1.bin doesn't exist.
is_whisper_model_present(large-v2)
is_whisper_model_present: returning false because ggml-large-v2.bin doesn't exist.
is_whisper_model_present(large-v3)
is_whisper_model_present: returning false because ggml-large-v3.bin doesn't exist.
curT0 = 0
curT1 = 309.296
Mode = transcribe
lang selection = french, params.language = fr
Enabling OpenVINO-based Encoder object that will run on CPU

very weird: before doing what you asked, i tried to launch audacity in administrator mode and transcription worked ONCE. In the same project, i tried a second time and it failed.
Then i ran audacity via admin cmd with the param you provided, and transcription failed again, the log you see is from this failure.

thank you for your help!

Hey @batmad, thanks! I didn’t see much from the log you shared… I was expecting to see the specific error after that last ‘Enabling’ message. But, you mentioned that it worked once and then failed on the second try… Which sounds very similar to a transcription issue that I debugged and will be posting a fix for soon (hopefully today). I think the next step here is to just download / install that updated patch release when it’s ready.

I’ll try to remember to comment here when it’s ready, but if you don’t hear from me, check here: Releases · intel/openvino-plugins-ai-audacity · GitHub

The release version will be 3.5.1-R2.1

Thanks,
Ryan

Hi @batmad,

The patched release is now available: Release v3.5.1-R2.1 · intel/openvino-plugins-ai-audacity · GitHub

Can you give it a shot?

Thanks!
Ryan

it works perfectly on this computer now!
many thanks!

This topic was automatically closed after 30 days. New replies are no longer allowed.