Illustrious XL v1.0 VAE - The High-Resolution Pioneer VAE
This is the official Variational Autoencoder (VAE) extracted from the landmark Illustrious XL v1.0 model by OnomaAI. It is the component responsible for the incredible clarity and detail that made v1.0 the first SDXL model to achieve native 1536px resolution.
If you want to imbue your generations with exceptional sharpness and high-resolution fidelity, this VAE is a must-have.
Key FeaturesEngineered for High Resolution: This VAE was co-developed with the main model to specifically excel at decoding large-scale images (up to 1536x1536). It is your best tool for reducing the common "hazy" or "washed-out" effect in high-resolution generations, ensuring crisp and clear results.
Clean and Foundational Aesthetic: Extracted from a "raw" pretrained base model, this VAE does not impose a heavy, specific art style. Instead, it provides a clean, clear, and detailed output, making it a highly versatile and compatible choice for a wide range of SDXL models. It enhances quality without overriding your chosen checkpoint's character.
Improved Detail & Clarity: By design, this VAE helps render fine details and textures with greater precision. Use it to bring out the subtle intricacies in fabric, hair, and complex backgrounds that might otherwise be lost or blurred.
True-to-Source Colors: It aims for a faithful color reproduction, avoiding oversaturation or strange color shifts. This makes it a reliable choice for achieving a balanced and professional look.
Place the file in your stable-diffusion-webui/models/VAE directory.
In the WebUI, select this VAE from the VAE dropdown menu (typically located in Settings -> Stable Diffusion).
Enjoy sharper, more detailed high-resolution images!
About The Illustrious XL Project
The Illustrious XL series is focused on pushing the boundaries of generative AI, with a core mission of achieving higher resolutions, robust prompt understanding, and providing powerful foundational models for creators.
Explore Our Models & Generate Online: illustrious-xl.ai
Check Out Our Full Platform: Discover exclusive models, high-res generation, and more on our official site.
Terms of Use: [Link to Terms of Use]
Your support helps us continue our research. If you like this VAE, please consider leaving a review and sharing your amazing creations!
Description
This is the first public, standalone release of the VAE from the foundational Illustrious XL v1.0 model.
We are releasing it as a standalone component to give the community access to the key technology that enabled the model's pioneering high-resolution capabilities.
Key characteristics of this VAE:
Source: Extracted directly from Illustrious XL v1.0.
Primary Function: To provide superior decoding for high-resolution images, specifically targeting resolutions up to 1536px. This was its primary design goal within the v1.0 project.
Aesthetic Profile: Clean, clear, and detailed. As it's from a base model, it is not aesthetically biased, making it an excellent general-purpose VAE for improving image quality.
Compatibility: Highly recommended for any SDXL 1.0-based checkpoint, especially when generating images larger than the standard 1024x1024.
Use this VAE to add sharpness and fidelity to your renders, carrying the legacy of the model that first broke the 1536px barrier in SDXL.