Character reference without LoRA training + 3D pose or regional layout control, in the same generation, for Anima Turbo (8 steps, CFG 1).

▶ Intro video (73 s): https://www.youtube.com/watch?v=j7wa5sC7VIU

Combines the Anima Reference Adapter or SnapRef v1b (character reference) with the Anima ControlNet-LLLite control models (3D pose, regional). Nodes and docs: https://github.com/quinteroac/ComfyUI-Anima-CharRef-LLLite

Workflows in the zip

• anima_charref_pose_simple — start here. Character reference + 3D pose from a mesh render on black. No Pixaroma or Pose Studio.

• anima_charref_posestudio_turbo — character reference + 3D pose from Pose Studio, optional pose import (SAM3D).

• anima_charref_posestudio_moss_upscale — the same plus an optional 2nd pass / hi-res fix with MOSS-Image up to 5 MP.

• anima_charref_regional_paint_turbo — 1 to 3 character references side by side + a region map painted in Pixaroma Paint.

• anima_snapref_* — the same three with SnapRef v1b instead of the Ref Adapter.

• anima_charref_snapref_* — both methods in one workflow; a switch turns on the Ref Adapter or SnapRef (one at a time).

Every workflow has a Power Lora Loader for your own Anima LoRAs (empty by default) and a note with every model and custom node it needs, with download links.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/ElectronGirl/ComfyUI-Anima-Ref-Adapter
git clone https://github.com/quinteroac/ComfyUI-Anima-CharRef-LLLite

Other custom nodes: comfyui-anima-safe-pag, comfyui-anima-style-nodes, rgthree-comfy, ComfyUI-Pixaroma (all but the simple workflow), ComfyUI_VNCCS_Utils (Pose Studio workflows), SnapRef node (SnapRef workflows).

Notes

• The LLLite control models were trained on Anima Base; with Anima Turbo they are experimental.

• Multiple characters: references side by side and a prompt that says who is on the left / middle / right. Two characters work best; three can mix features.

Credits: ElectronGirl (Ref Adapter), toyxyz (LLLite models), circlestone-labs (Anima), 0ppxnhximxr (SnapRef), AHEKOT (Pose Studio), pixaroma, rgthree, iljung1106, fulletLab, Comfy-Org / lodestones (MOSS-Image). Layout based on the Krea 2 Pose Studio workflow by foxfuressence.