Nice! I’m not too worried about performance on the CPU. For a lipsync system for example you would have some extra flexibility if you could edit the curves (audio) without worrying about the sound getting messed up. Basically the original lipsync line+a bunch of split phonemes that you could edit vs. just the phonemes which you couldn’t edit without messing up the sound of the line.
Animating lipsync by hand is just painful so this would be a way faster way to get this to work. If you use the actual sound half (or more) of the work is already done for you. I would just like a bit more artistic control which you could get by decoupling the actual sounds from the envelope follower curves.