Test the MotionDNA model
Describe a performance and the motion model synthesizes a movement preview — choreography, energy and stage presence from a single prompt.
landing/try.html and set
your CLERK_PUBLISHABLE_KEY (find it in the Clerk
Dashboard → API Keys, or run clerk auth login). The
publishable key is safe to expose in client code.
Sign in to access the playground
Access to the MotionDNA model is granted to approved accounts. Sign in or create an account to continue.
Prompt
Describe the artist, mood and movement.
When the MotionDNA inference server is connected, this stage plays back real skeletal motion generated by the trained model (live model). Without a server it falls back to an in-browser preview. Full video rendering is part of the closed beta — join the waitlist.
AI Music Video · HuMo-17B
Turn one photo + an audio clip into a lip-synced, audio-reactive performance video.
Pick a photo and an audio clip to begin.
The 17B model is loaded from disk first — fast when warm, but the first render after the GPU box wakes can take a few minutes to load. Then it denoises every frame for 30 steps (a Fast clip ≈ 8 min of compute on one H100). Keep this tab open while it renders.
Rendering runs the full 17B model on an H100 — a clip takes a few minutes. When no inference server is connected, this shows a sample render. Full pipeline access is part of the closed beta — join the waitlist.
AI Engine - FLUX - Qwen - LTX - Wan - ACE-Step
Generate images, video and music from the models installed on the H200 GPU.
Connect the engine and generate with the installed models.
Runs on the Boostify AI Engine (H200): FLUX / Qwen images,
LTX-2.3 / Wan 2.2 video and ACE-Step music. Point
window.ENGINE_API at your engine tunnel
(brev port-forward changing-purple-narwhal -p 8080:8080)
and set window.ENGINE_KEY.
Loading…
Preparing the MotionDNA playground.