If I record screen and audio when playing video games with my friends, will I be able to fine-tune an LLM+audio model on that dataset?

It'll be like that San Junipero episode of Black Mirror - immortality in a dark way.