Is it possible that “Dorico” will soon be using “Artificial Intelligence” in its software …?
“I’m afraid I can’t let you write consecutive fifths, Dave.”
What problems do you think that would solve, which Dorico’s existing algorithms don’t?
it depends on the marketing department
… to do or to assist/help you with exactly what?
First you need to define what artificial intelligence actually means in this context. AI can be anything, really, from a program that learns usage patterns and reacts accordingly (like an autocomplete function) to a script that composes music on its own.
You should rather ask which problems you want solved and then discuss the appropriate tools for the job.
I think that ranking search results in instrument section could get or is already AI’ed so that it shows your most used options. But apart from that better than implementing a more complex AI engine would be getting the scripting API to a level that one could implement an AI engine on top of it. This would be better for everyone I suppose.
Just another example: https://www.noteperformer.com/
Playback never sounded better
NotePerformer is the Artificial Intelligence-based playback engine for musical notation.
That is the claim! How do you guys think it sound?
Yes, NotePerformer is great as an AI, but the samples are pretty terrible… The small size of the app is great for laptop, or in case one doesn’t have
enough space for larger libraries. Would be nice if NP could be improved to use samples from Orchestral Tools, Spitfire Audio, CineSamples… etc.
Just like Staffpad does. I already asked Arne from Wallander Instruments for such improvement, and I hope he will do something to bring NP to next level.
The overall performance of NP is great and sounds pretty realistic, just the samples sound very unnatural.
Comparing NotePerformer with say VSL is a bit like comparing Joshua Bell playing a $500 fiddle made in a Chinese factory with a beginner playing a $10m Strad.
Different people may have different ideas about which “sounds better”, of course.
That’s because they are not samples.
Actually the sounds integrated in NP are samples. I know that from the creator of NP - Arne from Wallander Instruments.
The AI behind NP just controls the performance of the samples in order to bring more realistic sound.
The problem is that, Arne reduced the quality of the samples in order to make them to take a small amount of storage.
If you read Arne Wallander’s patent https://patents.justia.com/patent/20130287227 it is fairly clear that the quality of the original sample(s) is rather beside the point, compared with the processing done to it.
nevertheless, as Thurisaz says, it starts with samples and I bet that better samples would give more realistic sounding instruments even if it might make little difference to the way they phrase etc.