[gabrielgrant]

to be clear: you're calculating embeddings based on the timestamped transcript, yes? or are you doing something with the actual video content? what are you using for the embeddings?