So i was working on a small project, where i need to transcribe about 12 hours of audio. Now i used whipser 3 turbo to transcribe it, but now i have the problem of turning all that text into a summary or filtering stuff out. Now i was wondering if it's smarter to use a model specially trained for summaries or to just fire up llama 3.1 3b and throw the text at it?