How do I get whisper to allow larger files in the request?

Hello, I have been working on adding a feature to the app that allows the user to transcribe video and audios. After wiring everything up I get the following error. I am using NodeJS backend to basically make a call to whisper.

const whisperResponse = await this.openai.audio.transcriptions.create({
    model: "whisper-1",
    file: fs.createReadStream("audio.mp4"),
});

The error is as follows:

APIError: 413 Maximum content size limit (26214400) exceeded (26258790 bytes read)

I read somewhere about setting some values in the axios request but I am not using axios. I am using the OpenAIApi supported by Openai. I hope there is a way around this because the code works with tiny files but the user will expect to be able to transcribe large youtube, zoom and other mp4.

What do I have to do to fix this. Thanks.

You cannot fix what isn’t broken.

The user will expect you to turn their nonsensical huge files into reasonable data transfers - still with limits considering that you pay $0.10 per 15 minutes…

2 Likes

This is not a bug since this behavior is part of the documentation. The maximum file size is 25MB.

https://platform.openai.com/docs/guides/speech-to-text

@_j 's post above describes one way to deal with this situation. You can also split the audio file into smaller pieces.