Dev connects AI assistant to synth via MCP, uses spectrum feedback to iterate patches

sporeman456 · reddit · 2026-10-14

Max, developer of Ratio9, built a project that exposes a synthesizer's controls and audio analysis to an AI assistant over a local stdio MCP connection: the assistant can read/change parameters, pick wavetables, shape modulation, trigger notes, and request fresh measurements before revising a preset.

The core design distinguishes parameter proposals from evidence in the resulting audio: the spectrum shows frequency energy in the captured window, the spectrogram adds its movement over time — useful for inspecting a revision, but explicitly not a perceptual-quality score or a guarantee a patch works across registers and velocities.

The instrument stays manually usable, saved patches are independent of the chat, and measurement snapshots export separately rather than being embedded in preset files. The author notes this is a project explanation, not a benchmark; the demo video had waiting edited out and is not an uncut timing test.

Original post →

More from coding & agent

coding & agent channel →