Can this be used to remove retakes?

[orionflame]

Hi,

Basically I have some audio recording of myself for a video tutorial and I want to remove the retakes automatically. Here is a sample transcription from whisper-timestamped:

" Slurp Slurp or spherical linear interpolation is a term commonly used in computer graphics and animation, particularly in the context of rotations. Particularly for rotations."

As you can see:

particularly in the context of rotations

and

Particularly for rotations.

are retakes. So I want to keep the last retake but use the timing of the first one.

My idea is to use the hopefully accurate timestamps of words and pass this to Chat GPT to remove the redundant parts while keeping the timing of the first instance and use pydub to overlay/replace the earlier retakes and keep silence until the last retake.

Is this feasible or do you suggest another method? If the timestamps are accurate to edit the audio, it should be ok. I tried it manually using Chat GPT and it worked but I was only using plain text paragraphs from transcription but without any timing info. So I need to come up with a way that has both without confusing LLM. SRT subtitles didn't work for example, due to all the other noise present I think.

Thanks a lot in advance.