Illustrious XL 2.0 VAE - The Official VAE of the Illustrious XL Series
This is the official Variational Autoencoder (VAE) extracted directly from our flagship Illustrious XL v2.0 Base Model. It is designed to bring the signature visual fidelity and color science of the Illustrious XL series to your favorite SDXL models.
If you appreciate the clean aesthetic, vibrant colors, and high-resolution stability of Illustrious XL 2.0, this VAE is the key to achieving that look on other models.
Key FeaturesSignature Color & Tone: Imparts the renowned Illustrious aesthetic—rich, saturated colors, deep blacks, and excellent contrast without oversaturation. It's perfect for creating images with a clean, professional, and illustrative feel.
High-Resolution Fidelity: Engineered alongside the main model to perform exceptionally well at high resolutions (up to 1536px and beyond). It ensures crisp details and stable colors, reducing the washed-out or hazy look common in large-scale generations.
Reduces Artifacts: Helps to decode finer details more cleanly, mitigating common artifacts and improving the rendering of complex areas like eyes, hands, and intricate textures.
Broad Compatibility: As it's derived from a highly generalized base model, this VAE works exceptionally well with a wide range of SDXL 1.0 based models, from photorealistic to anime-style.
Place this file in your stable-diffusion-webui/models/VAE folder.
In the WebUI, select ixl-v2.0.vae.safetensors from the VAE dropdown menu in Settings -> Stable Diffusion.
Alternatively, most UIs allow you to select the VAE alongside the main checkpoint.
Note: This VAE will not change the composition or subject of your image. It exclusively affects the final color, lighting, and detail rendering during the decoding process.