{ "slug": "anymove-creature-mocap", "skill": "anymove", "upstream": "patch-gg/anymove@7a82a0a8fadffddd5c24fabf0697d9b6c814ce6e", "upstreamSha": "7a82a0a8fadffddd5c24fabf0697d9b6c814ce6e", "date": "2026-10-29", "engine": "Blender 5.2.3", "env": [ "Node v22.11.0", "Python 3.11 venv with the CPU build of torch 2.11.0 and numpy 2.4.6 (set AM_PY to its python); no NVIDIA GPU", "Blender 5.2.3 (set BLENDER); Pillow from the system Python for the page image", "No fal key, no downloads: the release's Tripo meshes, MiniMax clips, trained readers and LoRA are not used", "Not run: rigging a new mesh, filming, training a reader, reading a clip, the LoRA (they need CUDA or fal)" ], "trigger": "static", "passed": 23, "total": 23, "commands": [ "(cd examples && sh -c python3 tests/run_anymove_hands_on.py)", "(cd examples && sh -c python3 tests/run_anymove_hands_on.py)" ], "checks": [ { "criterion": "trigger", "name": "every skill in the pack says when to use it (6 skills)", "pass": true, "detail": "6/6" }, { "criterion": "trigger", "name": "triggers on: \"Animate my dragon creature from an AI video and export the animation as a GLB\"", "pass": true, "detail": "rank 1/118, overlap 6, via anymove" }, { "criterion": "trigger", "name": "triggers on: \"Motion capture a non-human monster from one generated clip onto its rig\"", "pass": true, "detail": "rank 1/118, overlap 7, via anymove" }, { "criterion": "trigger", "name": "triggers on: \"Train a joint reader for my spider creature so I can capture its moves\"", "pass": true, "detail": "rank 1/118, overlap 6, via anymove" }, { "criterion": "trigger", "name": "does not trigger on: \"Slice this sprite sheet into transparent frames\"", "pass": true, "detail": "rank 6/118, overlap 1, via anymove-rig" }, { "criterion": "trigger", "name": "does not trigger on: \"Write a Steam launch checklist for release day\"", "pass": true, "detail": "rank 5/118, overlap 1, via anymove-film" }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: mocap.solve fits the dragon's roar on the CPU (no GPU, no downloads)", "pass": true, "detail": "24 s" }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: the two views agree (reprojection under 10 px front, 12 px side; paw agreement score at least 0.85)", "pass": true, "detail": "front 5.93 px, side 7.99 px, paw 0.93" }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: motion.json covers every bone of the rig on every frame", "pass": true, "detail": "124 frames x 42 bones at 24.0 fps" }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: every rotation is finite and every root quaternion has unit length", "pass": true }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: the solved Euler angles stay inside the rig's joint limits (1 degree slack)", "pass": true, "detail": "0 values outside of 15252" }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: a stand-in armature rebuilt from rig_info.json matches its rest matrices", "pass": true, "detail": "STANDIN max_matrix_err=6.96161e-07 bones=42 verts=336" }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: blender/apply_motion.py keys the motion and writes animation.glb", "pass": true }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: Blender's joint positions match the solver's (under 2% of the creature's height)", "pass": true, "detail": "max 0.1 mm, creature 1.84 m" }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: animation.glb", "pass": true, "detail": "one skin with every bone, and an animation: 42 joints, animations ['rigAction']" }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: the clip keys every bone's rotation on every solved frame and lasts (T-1)/fps", "pass": true, "detail": "42 rotation channels x [124] keys, 1 moving translation (the root), 5.12 s" }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: Workbench renders of frames 0, 60 and 110 in both views exist", "pass": true }, { "criterion": "runs", "name": "hands-on: CPU fit of the bundled dragon roar, the repo's apply_motion.py on a stand-in rig, GLB checks: exits 0 with no script errors", "pass": true, "detail": "exit 0" }, { "criterion": "output", "name": "screenshot screenshot.png rendered", "pass": true }, { "criterion": "output", "name": "self-check list has at least 3 items", "pass": true, "detail": "6 items" }, { "criterion": "single", "name": "one job per skill (reviewed)", "pass": true, "detail": "One job split into steps: creature motion capture from an AI video. The router hands each step (rig, film, reader, capture, LoRA) to its own skill." }, { "criterion": "bilingual", "name": "English SKILL.md from upstream", "pass": true }, { "criterion": "bilingual", "name": "Chinese guide written", "pass": true, "detail": "1925 chars" } ] }