GPU rendering API
Real-time avatar rendering.
Pay with crypto. No accounts.
Upload a reference image, stream audio via WebSocket, receive lip-synced video via RTMP. Powered by FlashHead on A10G GPUs. Pay per minute with USDC on Base — no accounts, no API keys, just a wallet.
How it works
Create a session
POST /v1/sessions with a face image and RTMP output URL. First request returns 402 with USDC payment instructions. Sign and retry.
Wait for warm-up
Poll GET /v1/sessions/:id until status is "active". Cold starts take ~5 minutes. Warm starts are instant.
Stream audio
Connect via WebSocket. Send PCM audio frames (int16, 16kHz, mono). The GPU renders lip-synced video in real time.
Watch the output
Video streams to your RTMP endpoint. View in OBS, VLC, or pipe it anywhere. Extend the session or end it when done.
Integrate
Point your agent at the skill file. It contains everything needed to create sessions, stream audio, and render video.