OpenVINO transcription fails after loading AI model

Audacity 3 Help Forum
,
1

Hi there.

I installed the OpenVINO tools. I have tried selecting some audio for it to transcribe, with the default settings, and after it resamples and attempts to load the whisper AI model, it will have an error saying “Whisper transcription failed”.

I get the following output from the log.

  1. 19:29:22: 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. 19:30:29: Error: In Whisper Transcription Effect, exception: whisper.cpp context creation / initialization failed 19:30:30: Operation ‘OpenVINO Whisper Transcription’ took 66.467000 seconds. Poll was called 16947 times and took 1.454510 seconds. Yield was called 1144 times and took 0.590875 seconds. 19:30:28: Error: whisper_init_from_file_with_params(C:\Program Files\Audacity\openvino-models\ggml-base.bin, …) failed

I’m running Audacity v3.4.2.

Thanks in advance for any help. :slight_smile:

Device information

|Processor|11th Gen Intel(R) Core™ i7-11370H @ 3.30GHz 3.30 GHz|
|Installed RAM|16.0 GB (15.8 GB usable)|
|Graphics card|Intel Iris(R) Xe Graphics|

Edition Windows 11 Home
Version 22H2
Installed on ‎26/‎11/‎2023
OS build 22621.2861
Experience Windows Feature Experience Pack 1000.22681.1000.0
2

Hi @yassie_j,

Can you double check that you installed the ‘openvino-models’ folder in the correct place? More specifically, is the following file present on your system?

C:\Program Files\Audacity\openvino-models\ggml-base.bin

Thanks!
Ryan

3

Hi Ryguy,

Thanks for your reply. I followed the instructions in here exactly, and I downloaded the zips straight from there too. Releases · intel/openvino-plugins-ai-audacity · GitHub

I have checked the folder and that file is not there.

image
image1125×623 47 KB

I went back to the original zip file, which I downloaded straight from the github, and it’s not in the zip either.

Kind regards,
Yaseen

4

Here’s the zip file, with the file missing.

image
image1125×623 42.7 KB

5

Hmm, I would try downloading openvino-models.zip again. The contents should look like this:

image
image1391×1443 131 KB

BR,
Ryan

1 Like
6

Also, you can double check that zip file downloaded correctly by calculating md5sum (a checksum). If you open up and powershell window, and run this command:

certutil -hashfile openvino-models.zip MD5

You should get:

MD5 hash of openvino-models.zip:
04a239e3b04419460582cfe76f17467b
7

Hi Ryan,

Thank you so much!! I redownloaded the zip file and replaced the existing folder. I just did a test sample on a few seconds of audio and it transcribed it perfectly.

Take care,
Yaseen

1 Like
8

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