New tool2026-10-27via GitHubAI Video Generation

anymove

Name: anymove
Author: Patch (patch.gg)

anymove is an open-source Python pipeline from Patch that turns one AI-generated video of a rigged creature, filmed from front and side in the same frame, into a fitted skeletal animation exported as a GLB file, including wings, tails and eight legs.

Visit anymove ↗Repository ↗Release notes ↗

Free (open source) + paid APIsApache-2.0Python (3.11, needs Blender 5.x and a fal API key)★ 278 (as of 2026-10-28)by Patch (patch.gg)

Why it matters

For indie game makers and 3D artists who need animations for non-human creatures, where real motion capture doesn't exist. It replaces hand keyframing with an AI video clip plus an automated joint reader and rig fit.

Highlights

A video model (MiniMax H3 Max on fal in the examples) films the creature from two cameras in one frame; the two views pin each joint in 3D.
A DINOv2-based joint reader is trained on each creature from about 2,200 random poses; the README reports its average miss dropping from 40 px to 6 px for the dragon.unverified
The rig is fitted to the joints frame by frame and exported as a 24 fps GLB with no hand edits; the makers say reading takes about 10 s and fitting under a minute per clip.unverified
Ships as a Claude Code plugin with step-by-step agent skills (rigging, filming, training the reader, capture, LoRA training) that other agents can also load.
Example creatures (wolf, dragon, giant spider, fire demon) and trained readers are in the examples release.

Taken from the project's own README, docs or launch page. Claims about speed or accuracy are the maker's unless stated otherwise.

Try it if: You're building a game with dragons, spiders or other creatures and can't afford to keyframe every move by hand.

Worth knowing: Skeletons are still placed by hand, and the README notes video models sometimes drift or swap limbs. MiniMax H3's own community licence (not Apache-2.0) applies to its weights and the bundled wolf LoRA, and hosted use goes through fal's terms.

Quick facts

Pricing	Free code; per the README, about $0.40 per mesh (Tripo H3.1) and about $0.26 per 5-second clip (MiniMax H3 Max) on fal, plus about $12 to train the optional LoRAnot verified
Free or paid	Free (open source) + paid APIs
License	Apache-2.0
Language	Python (3.11, needs Blender 5.x and a fal API key)
Platform	Linux/desktop with an NVIDIA CUDA GPU (24 GB to train a reader; less to read and fit); one demo fit runs on CPU
GitHub stars	★ 278 (as of 2026-10-28)
Maker	Patch (patch.gg)

Related ranking

See how established tools compare in Best AI Video Generation. New releases aren't scored until we review them against the rubric.

What we couldn't verify

highlights[1] (reader accuracy is the makers' own result)
highlights[2] (timings are the makers' figures)
quick_facts.pricing (fal prices as of October 2026, per the README)

Checked 2026-10-28 from the tool's own pages. Prices and star counts change, so confirm on the official site. Listing here is not a ranking or an endorsement.