Illustrious XL

Illustrious XL is a powerful model based on Stable Diffusion XL (SDXL), developed by the OnomaAI Research team and specifically optimized for illustration and anime-style image generation. It inherits the excellent features of Kohaku XL-Beta - Revision 5 and improves on it to provide higher quality image generation capabilities.

Core features

• Based on SDXL: Inheriting the strong foundation of SDXL, it has higher resolution and stronger detail expression.

• Focus on illustration and animation styles: It has been specially trained and optimized for illustration and animation art styles, and can generate high-quality images of these styles.

• Open source: As an open source project, Illustrious XL encourages community participation, collaboration and knowledge sharing, aiming to promote innovation and development in the field of AI image generation.

• Technical documentation: Detailed technical documentation is provided for researchers and developers to gain an in-depth understanding of the technical details of the model. (Technical specifications can be viewed at https://arxiv.org/abs/2409.19946 )

Instructions for use

• Encourage open use: Illustrious XL encourages non-commercial use for research, learning, and creation.

• No closed source and commercialization: It is strictly prohibited to use this model for closed source projects or commercialize it in any proprietary way, which defeats the original intention of open source projects.

• Respect OnomaAI's contributions: This model was developed with resources provided by OnomaAI. The community is asked to respect their contributions and work together to maintain an atmosphere of open cooperation.

How to use Illustrious XL models in SDXL

1. Download the Illustrious XL model:

Download the .safetensors or .ckpt file for the Illustrious XL model on this page.

2. Place the model file:

Place the downloaded model file into the models/Stable-diffusion folder of Stable Diffusion WebUI. The full path to this folder is usually:

• Windows: stable-diffusion-webuimodelsStable-diffusion

• Linux/macOS: stable-diffusion-webui/models/Stable-diffusion

3. Select the model in AUTOMATIC1111 WebUI:

• Start or restart AUTOMATIC1111 WebUI.

• In the upper left corner of the WebUI interface, there is a drop-down menu for selecting the Stable Diffusion model.

• In the drop-down menu, you should see the Illustrious XL model you just placed. Select it.

4. Use Illustrious XL for image generation:

Now you can use Illustrious XL like any other Stable Diffusion model.

• txt2img (Text Image): In the "txt2img" tab, enter your prompt word (Prompt), and adjust other parameters (such as sampling method, number of steps, CFG ratio, etc.) as needed. Then click the "Generate" button to generate the image.

• img2img (图生图): In the "img2img" tab, upload an image as a basis and enter the prompt word. Adjust parameters such as "Denoising strength" and click the "Generate" button to generate a new image based on the uploaded image.

Tips and suggestions for use

• Prompting: Since Illustrious XL is optimized for illustration and anime styles, you will usually get better results using prompting words related to these styles. For example:

• masterpiece, best quality, 1girl, long hair, blue eyes, detailed face, beautiful background

• anime style, portrait of a young woman, intricate details, vibrant colors

• Negative prompt: Using negative prompts can help exclude unwanted elements and improve image quality. For example:

• low quality, bad anatomy, deformed, blurry, extra fingers, mutated hands

• HD fix (Hires. fix): Use the HD fix function to improve the resolution and details of the generated image.

• Sampler: Different samplers may produce different effects. You can try different samplers such as DPM++ 2M Karras, Euler a, etc. to find the one that best suits your preferences and prompt words.

• CFG Scale: CFG Scale controls how well the model follows prompt words. A higher CFG ratio will make the model follow the cue words more strictly, but some creativity may be lost.

• Experiment and tweak: Try different combinations of parameters and watch the results change to find the settings that work best for you.

How to get more information

For more information about the model and the latest updates, please see OnomaAI’s official release page.