Option to save and continue from a working in progress file for long audio

[nlgtuankiet]

Problem

Using whisper to process lenghthy audio (ex: > 1h podcasts) on an interruptable VM.
whisper only write the output file when it finishes processing the whole file. If the process is interrupted then all previous progress is lost and we have to process the file from the beginning next time hence computing resources are wasted.

Solution:

Add an option to specify the "work in progress file", whisper will write the result to that progress file or pick up the "progress file" from start working from there.

Ps: I'm new to python and incapable of implementing this at the moment. Python experts, please help us 🙏. This will save a lot of computing resources otherwise be wasted.