Desktop AI Robot for Developers
Deskman is a hackable desktop robot for developers, makers, AI researchers, and roboticists — a compact physical AI platform you can talk to, program, and completely rebuild.
Just about anything. You get a ready-to-program physical robot instead of a pile of dev boards, servos, cameras, speakers, and cables — log in and start building.
Local AI
Powered by an NVIDIA Jetson Orin Nano 8GB with CUDA acceleration and a 256GB SSD. The full conversational loop runs locally — no API connection required. Swap in any model you like.
As open-source AI improves, Deskman improves with it.
Deskman isn't a sealed consumer assistant — it's designed to be modified. The software runs on Linux and Python, and the hardware is accessible, so you can change how it behaves, replace models, modify the motion system, add sensors, or use it as the starting point for your own robot. The body is 3D printed, so it's easy to remix mechanically too.
$ ssh deskman@deskman.local
deskman@deskman.local:~$ cd speak
deskman@deskman.local:~/speak$ ./talk.py
Loading whisper...
Loaded on GPU in 5.7 sec
Loading kokoro...
Loaded on GPU in 10.0 sec, voice bm_fable
Loading text model server...
Started in 7.5 sec on GPU. Model: gemma-4-e2b
Starting text model inference...
Started in 4.5 sec
Follow-up open 20s
After Hi, talk for 20s without "robot". Then say "robot" to talk again. "robot quit" or CTRL-C to stop.Three actuated degrees of freedom on Waveshare bus servos. The head moves in pitch and yaw to turn toward people and react during conversation. A third actuator moves the hat, covering and revealing the camera and display. Drive it all from your own Python.
An Arducam camera gives Deskman sight — face tracking (out of the box), person and object detection, and visual interaction. The software pairs vision with the actuated head, so Deskman physically orients toward the people it sees.
A real, hackable robot that lives on your desk — or wherever you put him.
A couple of clips of Deskman doing its thing.
Software
Runs on Linux and Python, with CUDA accelerating speech-to-text, the LLM, and text-to-speech. Compatible with modern open-source STT, TTS, computer vision, and local language models — swap components as new models arrive. You're not locked into a proprietary AI service.
Use the models you want. Modify the code. Build something weird.
Talking to AI on a screen is useful. Talking to a robot sitting on your desk feels different. Deskman gives AI a physical presence — it can turn toward you, make eye contact, speak naturally, and serve as a platform for experimenting with embodied AI and human-robot interaction.
Deskman arrives assembled and ready to develop on.
