message: "Invalid file format. Supported formats: ['m4a', 'mp3', 'webm', 'mp4', 'mpga', 'wav', 'mpeg']"

[Ibby-debug]

I'm getting this error back while trying to record on a IOS using expo-av; I posted a question on stackoverflow with more details; Would appreciate any help or direction. https://stackoverflow.com/questions/76367660/react-native-using-expo-av-ios-mp4-file-openais-audio-transcriptions-invalid-fi

[ryanheise]

That error message is not a result of whisper, as whisper supports a much larger set of audio formats recognised by ffmpeg. I think you might be asking about Open AI's online API which wraps an endpoint around whisper but it provides a more limited set of features than whisper itself. In any case they do have their own discussion forum over there if you have a question about their service.

[benjieperez]

I also experience this using different mp3 format but returns invalid file format error.

This was a chunk of audios in 2 mins segments, sometimes its working sometimes not.

Done Transcribing temp/648fd37645c1a2374c099eb5_1.mp3
Done Transcribing temp/648fd37645c1a2374c099eb5_120001.mp3
Done Transcribing temp/648fd37645c1a2374c099eb5_240001.mp3
Retrying Transcription: temp/648fd37645c1a2374c099eb5_360001.mp3
{'error': {'message': "Invalid file format. Supported formats: ['m4a', 'mp3', 'webm', 'mp4', 'mpga', 'wav', 'mpeg']", 'type': 'invalid_request_error', 'param': None, 'code': None}, 'process_time': 2.6553714275360107}

[serhii-bilyk]

I have same issue on iPhone.

[ilyak1990]

For me specifically it was on iPhone, I was saving a valid .wav file (was working when I tested it) then I used a file type detector tool to find out it was actually some other file format that apple was saving it to, you can either convert to and from using node library ffmpeg or for ihpone specifically save it as a .m4a file instead of .wav

[Nolanjue]

if anyone else is still going through this issue, i think it might be data corruption from passing the audio file to the server. This discussion had a solution: https://stackoverflow.com/questions/77602701/openai-whisper-api-invalid-file-format-on-aws-lambda