How could I use whisper to transcribe long video locally on whisper V3?

[jasonsu123]

Hello everyone, I currently want to use Whisper for speech synthesis in videos, but I’ve encountered a few issues.

I am a Plus user, and I’ve used the paid API to split a video into one file per minute and then batch process.

But later I found out that the latest Whisper V3 can only be used locally and cannot be called via API?

I have some knowledge of Python programming. If I want to write a program on my computer, do you have any examples I can refer to?
My speech video is approximately 40 minutes long and over 500MB. I used the Whisper API before, but due to file size limitations, I had to split it for batch processing.

If I use Whisper locally on my computer, will there be no file size limitations?

Also, how can I find the names of different model types to use in the code, such as base, medium, large, and the latest V3?

Thank you.

[glangford]

To install whisper locally, start with the README document here which contains the information you need and the names of the models.
https://github.com/openai/whisper/blob/main/README.md

For local processing of a 500MB file, there is no need to split it up. Just provide the name of the file to whisper in the command line. As long as ffmpeg can process the file, whisper should be able to process the audio.

If you have questions, be sure to search the past discussions for any hints and check out whisper --help.

[mjyoussef]

Just as a follow-up, does model load the entire file into memory before transcribing or does it using streaming?

[ryanheise]

It loads the entire file into memory, although there is a PR to load in chunks here: #2256