Jussy Anima Workflow
An autocomplete-parameter ComfyUI workflow for Anima models.
📌 V4
✨ New Features
✅ Reworked Internal Pipeline
✅ Rework the Interface
✅ Reworked the LLM System Prompt
✅ Injected Separetly Quality Tags from LLM
Text Generation
A small language model helps you improve your positive prompt for Anima
Qwen3-VL-4B-Heretic reworks your prompt.
Preserve the model’s quality tags
Understand booru tags
Understand natural language
Convert into a language adapted to Anima
Organize the prompt
Polish and improve the wording
Add creative elements when requested
Supports NSFW prompts
Download and Place in ComfyUI\models\text_encoders
4 Models preloaded
BASE ANIMA, DIVING FLAT, RI-MIX and ANIMIJ-AI
Like always, you can modified all models and parameters
Positive Final Prompt
Displays the exact prompt sent to the model — including recommended quality tags, LoRA triggers, and your own input. If Text Generation is enabled, this shows the rewritten version produced by the LLM, which can differ from what you originally typed. It is saved in the image metadata.
Reduced Custom Node Requirements
ComfyUI-DaSiWa-Nodes is no longer required.
Unpacked Version
Due to recent ComfyUI updates, it is currently no longer possible to expose certain parameters.
Until this functionality becomes available again, I am providing an unpacked version of the workflow.
Required Custom Nodes
Download and Place in ComfyUI\custom_nodes
Required VAE and CLIP
VAE : qwen_image_vae
Download and Place in ComfyUI\models\vae
CLIP : qwen_3_06b_base
Download and Place in ComfyUI\models\text_encoders
How to Use
This workflow is pre-configured Anima Models.
To add your own checkpoint:
Open the Model Selector subgraph..
Select your checkpoint and change it inside the Model Selector..
Copy the recommended positive and negative tags from the model's Civitai page.
That's it!
The workflow will automatically apply the correct embeddings when the corresponding checkpoint is selected.
No need to manually copy the recommended quality prompts anymore. The workflow applies them automatically based on the selected model.
By default, Anima Base is selected and uses the official recommended embeddings.
To switch to another model, simply enable the corresponding checkpoint in the Model Selector.
Resolution Selector
The default resolution is 1344 × 1344 (≈1.7 MP).
You can change it in the Resolution section.
Upscale Model and Second Pass
I use 4x-AnimeSharp.
Download and Place in ComfyUI\models\upscale_models
You can still override parameters manually in the Enhance section if desired
Face Detailer
I use face_yolov11n.
You can change it in the Face Detailer section by your favorite.
Download and Place in ComfyUI\models\ultralytics\bbox
You can still override them manually in the Face Detailer section if desired
This is the Fourth Public Release. Feedback and suggestions are welcome.
V3
✨ New Features
✅ Added a Text Generation optional switch
✅ Added a Fourth Model selector
✅ Added Final Prompt visualizer
✅ Merged Upscale and Second pass
➖ Reduced Custom Node requirements
✅ Improved Post-Processing controls
📦 Added an Unpacked version
This is the Third Public Release. Feedback and suggestions are welcome.
V2
✨ New Features
✅ Added 2nd Pass optional switch
✅ Added automatic Steps per model
✅ Added automatic CFG per model
✅ Added automatic Sampler per model
✅ Added automatic Scheduler per model
⚠️ Bug fixed: Metadata of the upscale resolution not working
Each model now automatically uses its official recommended generation settings.
This is the Second Public Release. Feedback and suggestions are welcome.
V1
✨ Features
✅ Automatically loads Anima checkpoints and recommended quality tags
✅ 3 preloaded models
✅ Easy-to-use interface
✅ Optional Face Detailer
✅ Optional Latent Upscale
✅ Optional Save Image with Metadata
✅ Automatic file naming and folder organization
✅ LoRA Loader and Manager
✅ Ratio Selector for portrait and landscape format
This is the First Public Release. Feedback and suggestions are welcome.
Description
V3
✨ New Features
✅ Added a Text Generation switch
✅ Added a Fourth Model selector
✅ Added Final Prompt visualizer
✅ Merged Upscale and Second pass switch
➖ Reduced Custom Node requirements
✅ Improved Post-Processing controls
📦 Added an Unpacked version