Hi everyone,
I’ve been testing a camera-to-OSC workflow where a recognized hand gesture or body pose becomes a control signal inside TouchDesigner. The recognition runs on the source device and sends a configurable OSC address/value over the local network.
A simple TouchDesigner setup is:
- Receive the address with OSC In CHOP.
- Isolate the gesture channel with Select CHOP.
- For a one-shot cue, use Logic CHOP to react only to the off-to-on transition.
- For a toggle, feed that pulse into Count CHOP and wrap the result between 0 and 1.
- For a continuous control, keep the value stream and smooth it before exporting it to a parameter.
The part I’m currently tuning is false-trigger protection. On the recognition side I use similarity thresholds, hysteresis and configurable ON/OFF timing. In TouchDesigner, an additional short hold or cooldown can keep a pose from firing the same cue several times while the performer is moving into position.
I’d be interested to hear how others structure this in show files. Do you normally keep the debounce logic at the OSC source, inside TouchDesigner, or both?
For disclosure: the sender I’m developing is Magic Event Live. It has iPhone/iPad and native macOS editions, with limited free modes for testing the network workflow:
If useful, I can follow up with a small example network showing the momentary, toggle and continuous patterns.