Use AI as a show director behind safe TouchDesigner mappings

What it does

A Derivative forum thread proposing an AI layer that picks pre-approved cues and moods. Hazardous effects stay operator-only, and changes go through dry runs and approval queues.

How to set it up

1. Define a list of pre-approved cues and moods in TouchDesigner; this is all the AI may request.
2. Route AI output through a policy step: show intent, then a policy decision, then an approval queue or dry run.
3. Keep lasers, moving heads, blackout, PA and audio routing blocked or operator-only.
4. Rehearse with a simulated DMX output before connecting any real fixtures.

Source: Derivative forum

FAQ

Is this a finished product?

No. It's a concept thread, and the related tdmcp docs call the implementation a rehearsal proof of concept.

Who is it for?

Small clubs and parties without a pre-produced show, where an operator wants help adapting visuals live.