True Anime Palette
Community-trained on native Illustrious latents, ensuring perfect colour harmony and eliminating the muddy tones found in generic ControlNet models.
Precision Engineering for Anime Illustration
A dedicated ControlNet suite fine-tuned for the Illustrious-XL model. Achieve pristine colours, razor-sharp lines, and effortless 4K upscaling with unparalleled control.
Explore Models & DownloadWhy a Dedicated ControlNet?
Superior Line Art Fidelity
Our SoftEdge and Tile models respect ultra-fine outlines and intricate details, making them the ideal choice for professional manga inking and coloring workflows.
Efficient High-Resolution Renders
The Tile preprocessor splits the canvas for rendering, enabling you to create poster-sized 4K+ art on mid-tier GPUs with as little as 4GB of VRAM.
Model & Download Hub
Tile ≈ 2 GB
Intelligently upscales and extends your creations without sacrificing detail. Ideal for generating 4K+ artwork and seamless outpainting.
Download Tile Model (.safetensors) →OpenPose ≈ 2.3 GB
Transfer human poses with high fidelity. Perfect for controlling character anatomy, action shots, and complex multi-character scenes.
Download OpenPose Model (.safetensors) →SoftEdge ≈ 3 GB
Translates edge maps into clean, colored illustrations. Preserves delicate line art, making it essential for coloring sketches.
Download SoftEdge Model (.safetensors) →Depth ≈ 4.5 GB
Adds realistic depth, lighting, and perspective based on a depth map. Control the 3D space and atmosphere of your illustrations.
Download Depth Model (.safetensors) →
Installation: Place each .safetensors file inside your extensions/sd-webui-controlnet/models/ directory and restart the UI.
30-Second Quick-Start Guide
- Set Base Model: Choose Illustrious-XL as your Stable Diffusion checkpoint.
- Enable ControlNet: In the ControlNet panel, enable the unit and upload your reference image (e.g., a pose for OpenPose).
- Select Model: Choose the corresponding Illustrious ControlNet model (e.g., Tile, OpenPose) from the model dropdown.
-
Adjust Settings: For best results, start with a
Control Weightof 0.9 and leave guidance start/end at default (0 and 1). - Generate: Press Generate and watch the magic happen!
Frequently Asked Questions
Can I mix these with standard SDXL ControlNet models?
This is not recommended. The Illustrious ControlNet models are trained on a different latent distribution. Mixing them with standard SDXL models can lead to unexpected color shifting, artifacting, and a general loss of the signature "Illustrious" style. For best results, use only the dedicated weights.
Are these models compatible with ComfyUI?
Yes. All Illustrious ControlNet models are standard .safetensors files and are fully compatible with ComfyUI. Simply place them in your ComfyUI/models/controlnet/ directory and load them using the appropriate ControlNet loader node.
What are the optimal preprocessor settings?
For most use cases, the default preprocessor settings for each model type will work well. For Tile, if you are upscaling, ensure the "Tile" preprocessor is selected. For SoftEdge, `softedge_hed` often provides a great balance of detail and clarity. We encourage you to experiment to find what works best for your specific image.
Why is the Depth model so large?
The Depth model contains more extensive information about 3D spatial relationships and lighting, which requires a larger model size to capture accurately. This allows it to generate more nuanced and realistic perspective and environmental effects in your final illustrations.