I had to do sample retrieval, decoding, native CoreAudio logic to get acceptable latency. A programmable synth would have made my life so much easier.
supersonic-scsynth in Pi 5 feels more fit now, it’s the first time I tried a JS frontend for a macOS app but if it lets me import this and use it so easily, it already paid its dividends.
One more thought: it might be worth checking in with people before running AI agents over their GPL code, especially in a creative coding environment.
Multiple presses spawn multiple threads which layer over each other - unless you’re using Live Loops in which case it triggers hot code swapping.
What was the code you were running? Happy to help.
We also have a friendly forum to ask questions and get help: https://in-thread.sonic-pi.net