An Idea for an Intelligent Sampler

[Timmy Crowne]

While I’m firmly in the DAW-improvements-first camp, there is one idea I’ve been thinking about that would get me excited for a new device: an AI-based sampling engine.

I was reading about some recent developments in AI image generation, such as those outlined here: https://blogs.nvidia.com/blog/2022/03/0 ... -art-demo/. I started to wonder if machine learning could be used to “hear” the difference between a single sample mapped across the keyboard, and a multi-sampled, round-robin, velocity-sensitive instrument. Of course, specific instruments respond uniquely across their ranges, but let’s say an analysis is done of 20-30 instruments representing different categories. Extract those characteristics as “instrument signatures” in a new rack device that allows a realistic instrument patch to be generated from a single sample, with the user choosing what kind of signature to apply (plucked, bowed, blown, struck, etc.)

Need a guzheng VST? Grab a single sample online, drop it into Labyrinth AI Sampler, choose one of the plucked signatures and play away. Now try one of the bowed algorithms…

I would honestly be thrilled to spend serious cash on something like that. Thoughts?

[Yonatan]

AI in this field would be great.

[Mattvank]

Hello,

I am not sure to open a new thread for this so please apologize that i use yours. I was very suprised when i truncate in Kong a sample(edit sample) it truncate the real sample and there is no possiblity to undo that process.