Voice in, model out
Gibber handles the software side of a Muzzle session: capturing speech, sending it to the model, and playing back responses without leaking your side of the conversation into the room.
Muzzle · Project
Gibber. (short for jibberish)
The software project inside Muzzle: capture, routing, and the interface that turns a contained voice setup into a usable daily workflow with AI—without the jibberish leaking into the room.
What Gibber does
Built for daily use
Not a demo script. A stack shaped around real workflows—long sessions, interruptions, switching models, and the messy back-and-forth of thinking out loud with AI.
Part of the Muzzle stack
Hardware keeps your voice contained. Gibber is the interface layer that makes that containment useful: routing, session control, and the glue between you and the model.
Status
Gibber is in early development alongside Muzzle. If you want to follow progress, contribute, or talk through the use case, reach out.