AnyMove

patch-gg/anymoveReviewed by AI

What it does

A set of scripts and Claude Code skills that make animations for rigged 3D creatures (wolf, dragon, spider, demon) without a motion-capture suit. A video model films the creature from the front and the side in one frame, a vision model trained on that single creature finds its joints in both views, and the pair of views places each joint in 3D so the rig can be fitted and exported as a GLB. For technical 3D and game people: you need Linux-style commands, Python 3.11, Blender 5.x, an NVIDIA graphics card with CUDA (24 GB to train) and a paid fal.ai key for the clips.

How easy it is to install

D · terminal, several tools or a specific graphics card · Linux

D on Linux, the only system the instructions fit: you clone the repository, build a virtualenv with Python 3.11, install a CUDA build of PyTorch, put Blender 5.x on your PATH, add a fal.ai key and run a chain of `tools/*.sh` shell scripts; training the joint reader wants an NVIDIA card with 24 GB. Skeletons are still placed by hand on grid views. One smaller step works without a graphics card: the README shows fitting the example dragon's roar on the CPU.

Coming soon: install in 1 click with the RepoPicks app.

Numbers on GitHub

Stars: 448on Oct 31, 2026
Downloads: 1,183314 this week
Growth: +448stars in 4 days

Stars over time

Facts

Runs on: Linux
You need to install: Python, NVIDIA graphics card, Other tools
To build from source: Python, NVIDIA graphics card
Disk space: 950 MB–4.3 GB · estimate
Accounts or keys: fal.ai
Cost: Free, API costs extra · The code is Apache-2.0. The paid parts are listed in the README ("Prices on fal as of October 2026"): about $0.26 for a 5-second two-camera clip, $0.40 for a mesh from Tripo, and about $12 for training the optional MiniMax H3 LoRA. A new move for an existing creature is about $0.52.
License: Apache-2.0 (Permissive)
Can replace: Move AI, DeepMotion
Last update: Oct 27, 2026

Risk signals (6)

AttentionOwner account created recently
Owner account created 23 days ago
automatic check
AttentionDependencies with known vulnerabilities
Pinned dependencies with known vulnerabilities (OSV): torch: GHSA-rrmf-rvhw-rf47; pillow: GHSA-22hg-r758-rrw9, GHSA-24rc-8hj2-v98c, GHSA-254f-3936-288h
automatic check
Attentionlicence of what you make is not only apache
README licence section: "MiniMax H3's open weights carry a community licence and hosted use on fal goes through fal's terms: check both before shipping what you make with them", and the released wolf LoRA "is a LoRA on MiniMax H3, so MiniMax H3's community licence applies to it, not Apache-2.0." The Patch name and marks are also excluded from the licence.
AI review
Infodownloads models and example files on first use
requirements.txt: the DINOv2 backbone `facebook/dinov2-with-registers-base` is "downloaded on first use"; `tools/fetch_examples.sh` pulls the release files ("13 MB mesh, 355 MB reader, 4 MB clip"), and PyTorch comes from `--index-url https://download.pytorch.org/whl/cu128`.
AI review
Infoauthor lists cases that do not work
README "limits.": eight-legged walking does not fit ("The two views disagree about which legs are up (0.36-0.49)"), lying down fails, "Video models drift" (plain MiniMax H3 swapped limb colours on 2 of 5 wolf clips), and the ground contact term only knows wolf- and dragon-style paw names.
AI review
Inforepository is days old
Created on 2026-10-27 with a single release (`examples-v1`, same day) and already 359 stars; the deterministic checks also flag the owner account as 23 days old.
AI review

Evidence is shown in its original language.

What Radar could not check

Whether it runs on Windows (with WSL) or on a Mac; the commands and shell scripts assume a Linux-style system and an NVIDIA card.
Real disk use after PyTorch, the DINOv2 weights and the example files land on your machine.
Whether the animation quality matches the README's figures on your own creature and your own mesh.
What the whole process costs on fal for a project with several moves, and whether the October 2026 prices still hold.
How much hand work placing the skeleton takes for a creature that is not close to the four examples.