Idea for using an LLM to generate fine-tuning dataset from transcripts - View it on GitHub