I’m trying to use an m4a audio file in my scenario that’s located on Google Drive to create a transcript of that file using Whisper, however, I get an error stating that my file format is invalid. My scenario includes checking for a file in Google Drive, downloading the file, then transcribing it through Whisper.
I finally had to use an online converter to convert the file from m4a to mp3 which enabled me to run the scenario successfully.
Possibly. Try the same scenario with an MP3 file. Try it locally (using a Python script or something) to see if the m4a file works. If so, then you’ll know. I think someone had a similar problem with Whisper earlier this week and they fixed it by changing the file format to MP3.
Yes, converting it to mp3 worked, however we’re trying to automate the complete process so asking a user to convert a file adds a manual step, and integrating it with yet another app to do the conversion adds overhead and possibly cost.
True, but I believe it’s a problem with WHisper, not Make. The last time I tried to Whisper I had that problem. I was using an MP3 file but the bitrate was too high and I had to reduce it.
Other developers are hitting the same snag:
There are various workarounds around the web. See if they fit the bill.
If not, I would log a bug report with the Whisper team on Github.