Comment by michaelkrem

This sounds like something you could use LLMs for as LLMs are quite good at removing bias in content. You might run into context length issues feeding the whole transcript in though .