Easy Use Anima Input

Anima is a DiT on Cosmos-Predict2, which means "load a checkpoint" is really three separate loads: a diffusion model, a VAE, and a Qwen3-0.6B text encoder. Easy Use Anima Input is the node that gathers all three choices plus your prompt data into one context socket that Anima AiO Generator can swallow whole. If the generator is the engine, this is the ignition key - and it's deliberately boring, which is the point.

What it does

It takes two kinds of input and bundles them:

• EASYUSE_ANIMA_PROMPT_DATA - the structured prompt dict from Prompt Studio Advanced v2.
• Resource names - which diffusion model, VAE, and CLIP to use.

The output is a single easy use anima input socket that carries the whole context. Wire that to the generator's easy_use_anima_input input and you never drag another wire across the canvas for model loading again.

Why it works the way it does

The important design detail is what this node does not do: it doesn't actually load the models. The tooltip spells it out - the context stores only serializable prompt/resource data, and the AiO Generator loads MODEL, CLIP, and VAE at execution time. That's why there are no model patches or Torch Compile flags living inside a custom dict socket, and why a workflow stays reproducible: everything on this node is plain strings and a JSON blob, safe to serialize into the workflow file.

The inputs you actually set

• unet_name - the ANIMA diffusion model, loaded with ComfyUI's UNETLoader. The default is anima-base-v1.0.safetensors, which is exactly the file Anima base v1.0 ships as.
• vae_name - defaults to pixel_space, the Qwen-Image VAE Anima pairs with.
• clip_name - defaults to qwen_3_06b_base.safetensors, the Qwen3-0.6B Base text encoder.
• clip_type - defaults to qwen_image, the CLIPLoader type core Anima uses. The dropdown has the full list of CLIPLoader types, but you almost never leave qwen_image.

Everything else is either self-explanatory or hidden. input_settings is a versioned JSON field kept serialized for future resource settings - leave it alone. If you switch models a lot, the two that matter are unet_name and clip_name; the VAE and clip type stay put for nearly all Anima work.

Installing it

Same pack as everything else here:

cd ComfyUI/custom_nodes
git clone https://github.com/n0va39/ComfyUI-EasyUseAnima
cd ComfyUI-EasyUseAnima && pip install -r requirements.txt

then restart ComfyUI. ComfyUI Manager finds it under "ComfyUI EasyUse Anima".

Common issues

• Missing model files. The node is a picker; the actual files have to live in the usual ComfyUI directories - diffusion models under models/diffusion_models (or unet), the VAE under models/vae, the text encoder under models/text_encoders/clip. If the dropdown doesn't show a model you downloaded, that's a file-location problem, not a node problem.
• Wrong CLIP type for a custom Anima variant. If you load a non-standard build and get odd embeddings, check clip_type before blaming the sampler.
• Nothing downstream. This node is only useful when something consumes its output; if you're using it standalone you've grabbed the wrong node - the older Prompt Studio nodes output plain strings instead.

That's it. Set the three model picks once, leave it wired, and forget about it. The pack's whole "AiO" flow assumes this node sits between the prompt editor and the generator, and it's the least interesting node on the canvas for exactly the right reason.