I recently made them work with meeting presentation transcript. Mistral 7b (tiny one) was enough to reliably extract quite a lot of information! - It could der

I recently made them work with meeting presentation transcript. Mistral 7b (tiny one) was enough to reliably extract quite a lot of information! - It could derive speaker names, based purely on how people called themselves in the conversation - Drew a mermaid sequence diagram that wasn't perfect, but wasn't a complete garbage either. With few back and forth corrections it was on-point. - Created a truly usable meeting notes It was a much better UX than having to hunt down relevant video section, watch it and force to focus despite a lot of filler communication. That works very well for those kinds of 30 min. upward meeting, where the real content is in the spoken words and slides just reitarate that. I was really pleasantly surprised how much I liked that. Also, some models can rate each speaker contribution :o)