ComfyUI Anima Edit LoRA

Compatibility node that lets Anima Edit LoRAs work with ComfyUI's built-in ReferenceLatent workflow.

ComfyUI's built-in ReferenceLatent workflow passes one or more reference latents through the standard reference_latents condition. Anima Edit LoRAs expect those references to be concatenated onto the DiT latent time axis:

target latent + ref1 latent [+ ref2 latent ...] -> predict target

This node adds that model patch without replacing your LoRA loader. Load the Anima Edit LoRA with a compatible LoRA/adapter loader first, then apply this patch before sampling.

Node

• Internal key: AnimaEditLora
• Display name: Anima Edit LoRA
• Category: anima/edit/reference

The internal key and display name intentionally match the user-facing goal: making Anima Edit LoRAs usable with ComfyUI ReferenceLatent workflows.

Usage

1. Clone this repository into ComfyUI/custom_nodes/.
2. Restart ComfyUI.
3. Load an Anima model with the normal ComfyUI loaders.
4. Apply your trained Anima Edit LoRA with a compatible loader.
5. Add Anima Edit LoRA.
6. Use ComfyUI's built-in ReferenceLatent node to attach one or more reference latents to the sampler workflow.
7. Send the patched model to KSampler.

The node supports both single-reference and multi-reference workflows. Each reference latent must match the target latent channel count and spatial size. Batch size 1 references are expanded to the sampler batch, which covers normal CFG batching.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/wochenlong/ComfyUI-Anima-Edit-LoRA.git

Test

python -m pytest tests -q