innermost47/ai-dj — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a live AI music performance tool like OBSIDIAN Neural. I want it to run as a VST3, AU, and standalone app, so producers can use it inside a DAW or on its own.

The main idea is simple: type a prompt, pick an AI music model, wait around 30 seconds, then get a loop back that can be played instantly. It should have 8 tracks, MIDI triggering, pages of variations, a small step sequencer, and smooth page changes that stay locked to the beat. Add prompt saving, sample saving, drag and drop, and a way to bypass prompt enhancement when the user already knows what they want.

Make it feel like a playable instrument, not a song generator. Include per track volume, envelopes, delay and reverb sends, crossfaders, tempo sync, time stretching, standalone transport, Ableton Link support, and settings for server URL plus API key. Use current docs online if needed.

Want more depth? Deep Reverse