100% voice to voice models may be able to tell the difference between voices, I think Assembly AI has a way to differentiate voices currently in transcription so that might be useful