One idea I have is to use back-translation. After generating the new language audio, feed it back into the transcription, and then have an LLM compare it to the

One idea I have is to use back-translation. After generating the new language audio, feed it back into the transcription, and then have an LLM compare it to the transcript from the original. Penalty for any thought/detail that is missing. If too bad, start from scratch.