Send output to GPT-esque model to clean the output

[afonsojramos]

Whisper is great, but it has some downsides in some situations. Sometimes dialogue is not great, sometimes it is the speaker that doesn't have great english, but when this happens, the output is rarely clear, and a good way to solve that would be to rewrite everything from whisper with OpenAI's API or some other alternative.

What are the solutions here other than the one above? What are the advantages/disadvantages of each one?