Reachy Mini Gets Real‑Time Voice Brain Using GPT Realtime 2 and Opper Tool Layer
Deploy the reachy-voice-realtime repo by installing Python 3.12+, cloning the repo, and running the provided setup to give the Reachy Mini real‑time voice and motion capabilities.
Clone the reachy-voice-realtime repo, install dependencies, and run the setup script to equip the Reachy Mini with real‑time voice and motion control.
Summary
At a recent event, a Reachy Mini robot from Hugging Face was equipped with a real‑time voice brain by integrating GPT Realtime 2 through the Opper framework. The new system lets the robot hear via its mic, see through its camera, speak through its speaker, and trigger motion tools in real time. The repository https://github.com/opper-ai/reachy-voice-realtime provides a web UI that displays the camera feed, transcript, and live tool calls, and supports 19 motion and perception tools such as emotes, head/antenna/body movement, camera control, and sound direction. The robot mimics user gestures—waving, nodding, head tilting—while speaking. The realtime client and tool layer are separate, allowing routing to any provider or a local/OS realtime model. Setup requires Python 3.12+, is MIT licensed, and can be run with a single line of code. The authors handed the robot back to the event organizer’s daughter, enabling her to interact with it conversationally.
Key changes
- Adds GPT Realtime 2 integration via Opper for real‑time voice processing
- Provides a web UI to monitor camera feed, transcript, and live tool calls
- Supports 19 motion and perception tools (emotes, head/antenna/body movement, camera, sound direction)
- Mimics user gestures such as waving, nodding, head tilting
- Separates realtime client and tool layer for flexible routing to providers or local models
- Requires Python 3.12+ and is MIT licensed
- Repo: https://github.com/opper-ai/reachy-voice-realtime
- Enables Reachy Mini to physically react while speaking