Illustrious XL is a cutting-edge, open-source SDXL base model developed by OnomaAI Research, heavily optimized for native 1536x1536 high-resolution anime art and hybrid natural language tagging. It breaks the 1024px resolution limit of standard SDXL models, offering unparalleled sharpness and detail without upscalers. Key features include:
- Native High-Resolution Synthesis: Generates flawlessly at 1536x1536 and non-standard ratios.
- Hybrid Prompting System: Combines natural English sentences with Danbooru tags for precise control.
- Stable Text Encoder Finetuning: Superior concept understanding and prompt adherence.
- Massive Specialized Dataset: Trained on Danbooru2023, recognizing thousands of characters and styles.
- Multiple Versions: v1.0 (1536px milestone), v2.0 STABLE (cosine annealing), v3.5 VPred (maximized dynamic range).
- Open Source: Released under CreativeML OpenRAIL-M / Fair AI Public License, free for personal and research use.
- Compatible with ComfyUI, A1111, Forge, Draw Things.
- Community Ecosystem: Thousands of LoRAs and derivatives available.
Use cases include AI illustration, character design, concept art, LoRA training, and fine-tuning for specific aesthetics.
