Problem with file transcription

How To
1

Hi everyone,

I need some help figuring out how to transcribe long audio recordings using Whisper (OpenAI) in Make.

I have meeting recordings that are over 2 hours long, so they often exceed the 25 MB API limit. I know Whisper in Make has this restriction, but I’m not sure how to proceed when a file is, for example, 3 hours long and over 100 MB. I’m thinking maybe it could be parsed or split into smaller chunks, but I’d still like to preserve the original structure and eventually combine everything back into one complete transcript.

Has anyone dealt with this or found a good workaround?

Thanks a lot in advance for any help! :folded_hands:

image
image1870×760 73.9 KB

image
image1808×1652 365 KB

image
image2678×1482 299 KB

2

Hey Krystof,

you can use ffmpeg to cut the file in pieces. You can host a script on pythonanywhere and call it via ssh or http to send it the file and have it cut in smaller chunks. Then when they are ready, send them to whisper and combine the outputs.