🎨 AniSee | Personal Anime Fine-Tune of Anima Preview3 Base

Full Fine-Tune • Clean Anime Aesthetics • Tag + Natural Language • Anima-Compatible

Diffusion Model + AIO • 1 MP Native • LoRA-friendly • ComfyUI-ready

✨ What is AniSee?

AniSee is a personal full fine-tune of CircleStone Labs' Anima Preview3 Base, retrained on my own curated dataset to push the model further into a cleaner, more focused anime aesthetic.

It is not a LoRA merge — AniSee is a full fine-tune with around 20K training steps. The LLM adapter was only very lightly co-trained, following the official Anima fine-tuning guidelines.

The goal is to keep everything that makes Anima a strong illustration base:

• Danbooru-style tags

• Natural language prompts

• Mixed prompts

• Full Qwen text encoder support

• Qwen-Image VAE

• Anima-compatible generation behavior

while shifting the default style toward a stronger, cleaner anime look in line with my other checkpoints.

AniSee is mainly intended for:

• Anime-style illustrations

• Character-focused images

• Cleaner anime aesthetics

• Style experiments

• Testing Anima-based fine-tunes inside ComfyUI

• Simple plug-and-play ComfyUI generation with the AIO version

🎯 Key Features

• ✅ Full fine-tune on Anima Preview3 Base — not a LoRA merge

• ✅ Around 20K training steps on a curated anime dataset

• ✅ Clean, focused anime aesthetics

• ✅ Supports Danbooru-style tags, natural language, and mixed prompts

• ✅ Compatible with the standard Anima ComfyUI workflow

• ✅ Drop-in replacement for anima-preview3-base.safetensors when using the diffusion-model-only version

• ✅ LoRA training friendly — same base architecture as Anima

• ✅ AniSee AIO version available

• ✅ AIO includes image model, Qwen text encoder, and Qwen-Image VAE in one checkpoint

• ✅ Loads directly through the standard ComfyUI Checkpoint Loader

• ✅ No separate text encoder or VAE download required for the AIO version

• ✅ Simple install: place the AIO file in ComfyUI/models/checkpoints/

• ✅ Only around 5.5 GB despite including the image model, text encoder, and VAE

• ✅ Smaller than many popular large anime checkpoints such as Pony, SDXL, or ILL-style models, while still being a complete all-in-one package

• ✅ Recommended version if you want the easiest setup and a clean plug-and-play ComfyUI experience

🖼️ Example Gallery

I put together a curated gallery with sample images generated by AniSee — characters, scenes, different styles and prompts in action.

Worth a look if you want to see what the model can do before downloading:

🎨 View the AniSee Sample Gallery →

🗺️ AniSee Roadmap

✅ Released

🎨 AniSee Base
Full fine-tune of Anima Preview3 Base, diffusion-model-only variant.

📦 AniSee AIO
All-in-one checkpoint with the AniSee image model, Qwen text encoder, and Qwen-Image VAE integrated into a single file.

This is now the recommended version for users who want the easiest setup.

🔧 Official AniSee ComfyUI Workflow
A dedicated workflow will follow, although the AIO version already works with the standard ComfyUI checkpoint workflow.

🔜 Planned

🚀 AniSee Turbo / 4-Step CDM
A future few-step version distilled with Continuous-Time Distribution Matching. The goal is fast 4-step anime generation with strong details and clean style retention.

More updates coming as testing progresses! 🎨

📦 Versions Overview

🟣 AniSee AIO — Recommended

This is the easiest and most convenient version of AniSee.

The AIO checkpoint includes everything needed in one file:

• AniSee image model

• Qwen text encoder

• Qwen-Image VAE

Place the file in:

ComfyUI/models/checkpoints/

and load it with the standard ComfyUI Checkpoint Loader.

No separate text encoder download.
No separate VAE download.
No searching for missing files.

Just put the checkpoint into the correct folder and load it like a normal ComfyUI checkpoint.

The AIO file is around 5.5 GB, which is still smaller than many popular large anime checkpoints such as Pony, SDXL, or ILL-style models — even though the text encoder and VAE are already included.

This is my personally preferred way to load and use the model: simple, clean, self-contained, and beginner-friendly.

🟢 AniSee Diffusion Model

This is the separate diffusion-model-only version.

Use this if you already have the Anima setup installed and prefer loading the diffusion model, text encoder, and VAE separately.

Required files:

• AniSee diffusion model

• Qwen text encoder

• Qwen-Image VAE

This version is useful for advanced workflows, custom loader setups, or users who want more control over each component.

🚀 Turbo 4-Step / CDM — Planned

Few-step distilled variant for fast generation.

Based on the CDM paper, Continuous-Time Distribution Matching for Few-Step Diffusion Distillation, this future version is intended to produce clean anime generations at very low step counts while keeping strong fine details.

💜 Why AniSee AIO?

The AIO version is made for users who want the simplest possible ComfyUI setup.

Many modern image models require multiple files, separate loaders, extra text encoders, VAEs, or manual workflow changes. AniSee AIO avoids that.

You only need one checkpoint file.

Put it into:

ComfyUI/models/checkpoints/

load it with the standard ComfyUI Checkpoint Loader, and start generating.

Even with the image model, text encoder, and VAE packed into one file, AniSee AIO is only around 5.5 GB. That makes it smaller than many large anime checkpoints while still being a complete all-in-one package.

For me, this is the cleanest and most comfortable way to use a model: no hunting for missing components, no confusing setup, no extra dependency chaos. Just load and generate.

🎨 AniSee Base — Recommended Settings

The settings I personally use and recommend as a starting point:

Steps:       40
CFG:         4.5
Sampler:     er_sde
Scheduler:   simple
Resolution:  ~1 MP, for example 1024×1024, 896×1152, 1152×896

CFG Guide:
4.0–5.0 is the sweet spot for balanced quality and creativity.

Going above 5.0 starts to risk burning the image, especially with heavy quality tags. If results feel too harsh, drop CFG slightly or reduce the quality tag count.

Sampler alternatives

All of these work well, just with slightly different character:

• er_sde + simple — my default, neutral style, flat colors, sharp lines

• euler_a — softer, thinner lines, slightly more 2.5D feel, tolerates higher CFG

• dpmpp_2m_sde_gpu — similar style to er_sde but more creative, can get wild on short prompts

Feel free to experiment — these are starting points, not hard rules.

📐 Resolution Guide

Use Case Resolution ⭐ Square / General purpose 1024 × 1024 Portrait / Character art 896 × 1152 Landscape / Scenes 1152 × 896 Wider cinematic 1254 × 836 Widescreen 1365 × 768

Stay around 1 MP for the cleanest results.

The Anima base starts breaking down somewhere around 2 MP, so if you want bigger images, generate at 1 MP first and upscale afterwards.

💡 Prompting Guide

AniSee inherits Anima's prompting system.

It accepts:

• Danbooru / anime-style tags

• Natural language prompts

• Mixed prompts with tags + sentences

A good prompt structure:

[quality tags] [meta tags] [safety tag] [subject] [character] [appearance] [pose] [clothing] [background] [lighting] [style]

Important tag rules inherited from Anima

• Use lowercase for tags, spaces instead of underscores

• Score tags are the only tags that use underscores, for example score_7

• Artist tags must be prefixed with @, for example @artistname

✅ Good Prompt Example — Mixed Prompt

masterpiece, best quality, score_7, highres, illustration, safe, 1girl,
long silver hair, blue eyes, black hoodie, standing in a rainy city street
at night, neon lights reflecting on wet asphalt, cinematic lighting,
detailed anime illustration

✅ Good Prompt Example — Natural Language

masterpiece, best quality, score_7, highres, illustration.
A young anime girl with long silver hair and golden eyes, wearing a
traditional shrine maiden outfit with white haori and red hakama.
She stands in a sunlit bamboo forest, cherry blossoms falling softly
around her. Warm afternoon light filtering through the trees,
detailed fabric shading, calm serene expression.

❌ Avoid

Very short tag dumps like:

anime girl, silver hair, hoodie

The model can produce unexpected results when the prompt is too sparse.

Aim for at least a few descriptive tags or 2+ sentences.

⭐ Recommended Positive Prefix

Start every prompt with:

masterpiece, best quality, score_7, highres, illustration,

Then add your subject, character, scene, and style tags after that.

You can also experiment with other quality tag combinations:

masterpiece, best quality, score_7, safe

masterpiece, best quality, score_8, highres, official art

score_9, masterpiece, absurdres, anime screenshot

But the prefix above is what I personally use and recommend as a starting point.

⭐ Recommended Negative Prompt

This is the negative prompt I run with — it cleans up most common issues without being so aggressive that it kills the style:

worst quality, low quality, score_1, score_2, score_3, artist name,
(lowres:1.2), (worst quality:1.4), (low quality:1.4), (bad anatomy:1.4),
bad hands, multiple views, comic, jpeg artifacts, patreon logo,
patreon username, web address, signature, watermark, artist name,
censored, mosaic censoring

If your images come out too flat or lose style, reduce the weights on the heavier terms, for example drop (low quality:1.4) back to low quality.

🛡️ Safety Tags

Inherited from Anima.

Use one of these in the positive prompt:

• safe — for normal generations, recommended default

• sensitive

• nsfw

• explicit

🔧 Installation

There are two ways to use AniSee.

Option 1 — AniSee AIO Version Recommended

This is the easiest setup and my recommended version.

Step 1 — Download the AniSee AIO checkpoint.

Step 2 — Place the file here:

ComfyUI/models/checkpoints/

Example:

ComfyUI/models/checkpoints/AniSee-AIO-v1.safetensors

Step 3 — Load it in ComfyUI with the standard Checkpoint Loader.

The AIO version already includes:

• Image model

• Qwen text encoder

• Qwen-Image VAE

You do not need to download or manually select a separate text encoder or VAE.

This makes AniSee very easy to use with the standard ComfyUI workflow.

Option 2 — Diffusion Model Version

Use this version if you want the classic Anima-style setup with separate files.

Place the files here:

ComfyUI/models/diffusion_models/
ComfyUI/models/text_encoders/
ComfyUI/models/vae/

Example:

ComfyUI/models/diffusion_models/AniSee.safetensors
ComfyUI/models/text_encoders/qwen_3_06b_base.safetensors
ComfyUI/models/vae/qwen_image_vae.safetensors

Then load them with the standard Anima workflow:

• Load Diffusion Model → AniSee.safetensors

• Load Text Encoder → qwen_3_06b_base.safetensors

• Load VAE → qwen_image_vae.safetensors

If you already use Anima Preview3 Base, you likely already have the required text encoder and VAE.

📈 Version History

v1.1 — AniSee AIO Release

• Added AniSee AIO checkpoint

• Image model, Qwen text encoder, and Qwen-Image VAE are packed into one file

• Loads directly with the standard ComfyUI Checkpoint Loader

• No separate text encoder or VAE download required

• Place the file in ComfyUI/models/checkpoints/

• Around 5.5 GB total size

• Recommended version for the easiest AniSee setup

• Keeps the same AniSee anime-focused full fine-tune style

v1.0 — Initial Release

• AniSee Base — full fine-tune of Anima Preview3 Base

• Around 20K training steps on a curated anime dataset

• LLM adapter only very lightly co-trained, following Anima's fine-tuning guidelines

• Diffusion Model variant

• Compatible with the standard Anima ComfyUI workflow

• Drop-in replacement for anima-preview3-base.safetensors

🙏 Credits

• Base Model: Anima Preview3 Base by CircleStone Labs and Comfy Org

• Underlying Architecture: Built on NVIDIA Cosmos-Predict2-2B. Anima is a derivative model.

• Fine-Tune: SeeSee21

• CDM Distillation Method, planned Turbo variant: Continuous-Time Distribution Matching for Few-Step Diffusion Distillation — Liu et al., 2026

📜 License

AniSee inherits the CircleStone Labs Non-Commercial License from Anima.

The model and derivatives are usable only for non-commercial purposes.

As a derivative of Cosmos-Predict2-2B-Text2Image, the NVIDIA Open Model License Agreement also applies insofar as it covers derivative models.

For commercial licensing of the base model, please contact CircleStone Labs at:

[email protected]

AniSee — a personal anime fine-tune of Anima Preview3 Base. 🎨