Anima Support

[Acly]

Anima

State: Basic generation, inpainting (fill/extend), upscaling, sketch/lines control layers

Update 1.52.0: regions are now supported.

Requirements

Requires plugin version 1.51.0 or later.

For custom ComfyUI installs remember to update:

• ComfyUI
• comfyui-tooling-nodes

Model Download

Models need to be downloaded manually. If you are using a plugin-managed installation, please locate the models folder in the place where you installed the server. Download the models to the appropriate sub-folder indicated below.

Overview

📂 models/
├── 📂 text_encoders/
│   └── qwen_3_06b_base.safetensors
├── 📂 vae/
│   └── qwen_image_vae.safetensors
├── 📂 diffusion_models/
│   └── anima-base-v1.0.safetensors
└── 📂 controlnet/
    ├── anima-lllite-any-test-like-v2.safetensors
    └── anima-lllite-inpainting-v2.safetensors

text_encoders

• qwen_3_06b_base.safetensors

vae

• qwen_image_vae.safetensors

diffusion_models

• anima-base-v1.0.safetensors

controlnet

• anima-lllite-any-test-like-v2.safetensors
• anima-lllite-inpainting-v2.safetensors

[KallyWally]

[caseyotaku]

[Miraihi]

Gives me error "Server error: ERROR: clip input is invalid: None
If the clip is from a checkpoint loader node your checkpoint does not contain a valid clip or text encoder model."
Local managed server.
I have text encoder, vae and checkpoint file downloaded and put into the right folders. Obviously the plugin's version is 1.50.0.

[Acly]

Error message sounds very much like anima-preview3-base.safetensors is in the checkpoints folder, which is not the right folder.

[Miraihi]

Alright, I guess I decided that it's not important since the model got recognized anyway. But turns out it's important. Sorry to bother, now it works.

[Miraihi]

By the way, what are a good combinations of sampler and scheduler for this kind of models? Euler A karras work fine but most of the presets I made for illustrious work poorly (Fried, blurry, or broken in other bizarre ways).

[Miraihi]

Addition: er_sde works, but only with simple scheduler. I have a hunch that the choice of scheduler is important here. More testing required.
It does work beautifully though!

    "ER_SDE": {
        "sampler": "er_sde",
        "scheduler": "simple",
        "steps": 30,
        "minimum_steps": 4,
        "cfg": 4.0
    },

[Miraihi]

Found a pretty good Turbo preset, check it out

    "ER_SDE Anima turbo": {
        "sampler": "er_sde",
        "scheduler": "simple",
		"lora": "anima-turbo-lora-v0.1.safetensors",
        "steps": 12,
        "minimum_steps": 2,
        "cfg": 1.5
    },

It can get away with 6 and less steps.

[Miraihi]

It seems like most samplers do well as long as you use the simple scheduler. I wonder what are the other scheduler options.

[Miraihi]

Schedulers tested with dpmpp_2m_sde, steps: 20, cfg: 6:

1. normal - Scrambled, probably usable with other samplers and other settings
2. simple - Good
3. sgm_uniform - Good
4. ddim_uniform - Fried, probably usable with other samplers and other settings
5. beta - Quirky, probably usable with other samplers and other settings
6. linear_quadratic - Quirky, probably usable with other samplers and other settings
7. kl_optimal - Quirky, probably usable with other samplers and other settings
8. gits - Black screen, Bad
9. polyexponential - Black screen, Bad
10. laplace - Black screen, Bad

[Miraihi]

Overall ER_SDE simple or Euler simple seem to be the most stable options.

[fink94]

Guys, Im a bit dense.
If the directories of models goes through yaml, won't it work?

other_ui:
  base_path: E:/sd-webui-forge-neo/
  checkpoints: models/Stable-diffusion
  diffusion_models: models/Stable-diffusion/Anima
  loras: models/Lora
  vae: models/VAE
  embeddings: embeddings
  controlnet: models/ControlNet
  upscale_models: models/ESRGAN
  hypernetworks: models/hypernetworks
  text_encoders: models/text_encoder

I cant get this to work:(

[Acly]

do not link checkpoints and diffusion_models to the same (parent) folder, keep them separate

[fink94]

Thanks a lot, Boss, everything works!:)
I also wanted to thank you for your work.
Thank you so much! You and your team are awesome.

[Apacchi88]

Anyone knows how to work with the Comfyui-Anima-LLLite, specifically the inpaint model in Krita?
I can work with it in normal comfyui interface, but don't know how to get it to work in Krita.

[Acly]

Inpaint LLLite is used automatically at 100% strength for Fill/Expand/etc. you don't have to do anything.
But you can choose "Generate (Custom)" option when you have a selection active, and toggle "Seamless" option manually. This toggles LLLite inpaint.

[Apacchi88]

Inpaint LLLite is used automatically at 100% strength for Fill/Expand/etc. you don't have to do anything. But you can choose "Generate (Custom)" option when you have a selection active, and toggle "Seamless" option manually. This toggles LLLite inpaint.

AH okay. Thanks I was trying to get it to work with Live mode, completely forgot to check Generate/Fill mode

[octoepictraveler-svg]

What am I doing wrong?

I've placed every file on the respective folder but Krita tells me I don't have the Anima Workload installed; there's no Anima workload to download on the plugin menu either.

Then when I try to generate, it says I'm missing the model in the checkpoint folder, but when I add the model on that folder, it generates more errors about several qwen text encoders not being installed...

[Miraihi]

You've made the same mistake as me - put the base model into the usual checkpoints folder. They go into the diffusion_models folder instead.

[octoepictraveler-svg]

You've made the same mistake as me - put the base model into the usual checkpoints folder. They go into the diffusion_models folder instead.

Even if I don't have the anima safetensor on diffusion_models and not on checkpoints, it still prompts me out with this, am i supposed to make a new style? remove all the models unrelated to anima from checkpoints?

[Miraihi]

You must put the anima .safetensors model into diffusion_models folder, that's it. Just like in the instructions in the post. Consider this family of models a special case.

[octoepictraveler-svg]

I seriously don't know what am I doing wrong

[Miraihi]

Looks fine, pretty much like my custom style.
Looks like an incomplete installation. You see what's missing - the text encoder. Re-download it in case it's corrupted for some reason.
If it fails maybe try to rollback to 1.51.0.