.wav files not working in OpenAI because of audio/x-wav instead of audio/wav mimeType

[barryvdh]

When I try to transcribe a .wav file with OpenAI, I get:

OpenAI Error [400]: invalid_request_error - Audio file might be corrupted or unsupported

If I override the mimeType to audio/wav, it works correctly.

$response = Prism::audio()
    ->using('openai', 'gpt-4o-transcribe')
    ->withInput(Audio::fromLocalPath(path: '/path/to/file.wav', mimeType: 'audio/wav'))
    ->asText();

This is because the Media.php detect it like this:

        if (! $mimeType && ! ($mimeType = File::mimeType($path))) {
            throw new InvalidArgumentException("Could not determine mime type for {$path}");
        }

which returns audio/x-wav.

I'm not sure what the difference is, but I do see that OpenAI doesn't accept x-wav. I don't know if we should override the Media mimeType, or the OpenAI provider handling of this.

As said, override the mimeType works, but I think more people will run into this