Text-to-image graph for Anima.
Write a prompt, set sampler / seed / cfg / size, run. After the base image you can hires, then face, then hands. Each stage is a toggle. Save is a toggle too. Off means skipped.
ControlNet is optional (toggle 00). It goes through Anima LLLite depth.
If you use ControlNet:
- models/model_patches/anima-lllite-depth-1.safetensors
- models/geometry_estimation/depth_anything_3_mono_large.safetensors
Face / hands wildcard:
- [CONCAT] text — append to the base prompt
- empty — same prompt as base
- anything else — replace the prompt on that stage
Plumbing you don't need to touch is inside subgraphs.
Description
drop anima_simple_clean.json into comfy. install whatever manager screams about. queue.
this is a graph, not a preset. sampler, size, steps, loras, which stages run — that's why the knobs are on the canvas. change them.
controlnet is anima lllite depth. not instantx. anima is qwen3-0.6b, they will not become friends.
hires / face / hands are toggles. off means skipped, not broken.