toyxyz open-sources dual Anima ControlNet-LLLite adapters for 3D pose and region control
bdsqlsz · x · 2026-10-13
toyxyz trained and released two lightweight ControlNet-LLLite adapters built on Anima Base 1.0, now on Hugging Face.
- anima3dpose6200: conditions on shaded 3D body meshes to control character pose, including fingers; 6,200 updates total, with pose and hand accuracy still variable.
- animaregional4335: uses a 3-channel RGB instance-color map where solid colors mark subject regions and black marks unassigned pixels, controlling layout, silhouettes and object placement — it is a region-conditioning adapter, not a mask-based inpainting model.
For multi-person scenes the two can be combined: the pose adapter guides body posture while the regional adapter defines distinct spatial regions for each character. Training used kohya's sd-scripts; usage is via ComfyUI.
Related event: toyxyz Open-Sources Two Anima ControlNet-LLLite Adapters(3 posts)→
More from coding & agent
- GPT-6.1 Sol Pushes PageSpeed Scores to Near-Perfect in Real Test — CtrlAltDwayne · 2026-10-13
- Dev dogfoods his infra visualization tool, saves $340/mo with one click — DanielLockyer · 2026-10-13
- Proxy gateway with semantic routing cut multi-step agent LLM costs by 54% in production — d13-m1z · 2026-10-13
- Non-dev marketer rebuilt his whole marketing stack with Claude Code and stopped buying SaaS — summit_23 · 2026-10-13
- Daniel Jeffries: only ~2% can actually use AI agents, mass adoption is 10-20 years away — Dan_Jeffries1 · 2026-10-13
- Fully local AI short film mimics Wes Anderson style with a detailed open pipeline — Mist3r-Robot · 2026-10-13