Humanoid Network¶
The motion data layer for Physical AI. Any Robot. Any Task. One Network.
Robots can't learn what they've never seen, and the motion data to teach them barely exists. Humanoid Network is where that data gets made. You generate a robot motion in a browser simulator, Hydra scores it against physics, and accepted motions join a verified library that Physical AI can train on.
No hardware. No GPU of your own. Everything runs from your browser.
What you can do today¶
-
Create Motions
Generate robot motions from approved prompts in the simulator.
-
Browse Motions
Explore proofed motions, Hydra scores, replays, and contributors.
-
Foundation Dataset
Inspect the 46,893-episode corpus that Hydra reasons against.
-
Get Started
Connect a wallet and generate your first motion in minutes.
How it works¶
- Create. Pick an approved prompt and generate a motion in the simulator.
- Score. Hydra checks the motion against physics and records quality signals.
- Proof. Accepted motions can receive an on-chain proof record (Base Sepolia during testing).
- Browse. Proofed motions enter the verified library for anyone to inspect.
The Foundation Dataset sits behind all of this: an independent corpus of 46,893 episodes, embedded in 3D behavior space, that gives Hydra the context it needs to score new motion well.
What's live today¶
Token features aren't live yet. The HAN token launches on Base at the token generation event (TGE), and the app keeps every token action switched off until that official contract is live and verified. Motions you proof now are recorded on Base Sepolia as rehearsal records: they prove the pipeline works, but they carry no mainnet value or allocation, and they may inform future protocol eligibility only once those features launch. When token features go live, we'll say so here.