I tend to agree with you, as I have spent many years in lessons myself and taught for a couple. But a) real lessons are expensive and difficult to access for a lot of people, and b) even for someone who has a teacher, most practice time is self-directed. I think there's a real opportunity to make practice much more efficient, and also provide a better experience than currently exists for those who can't access a human teacher.
Also, multimodal LLMs are improving so quickly that I wouldn't be that surprised if they were able to do what you're describing in a couple of years. But they are definitely not there yet.
Also, multimodal LLMs are improving so quickly that I wouldn't be that surprised if they were able to do what you're describing in a couple of years. But they are definitely not there yet.