<h3 id="illustrious-xl-2.0-vae-the-official-vae-of-the-illustrious-xl-series-158fg4hu7">Illustrious XL 2.0 VAE - The Official VAE of the Illustrious XL Series</h3>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 Features<ul><li>Signature Color &amp; 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.</li><li>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.</li><li>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.</li><li>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.</li></ul>How to Use<ol><li>Place this file in your stable-diffusion-webui/models/VAE folder.</li><li>In the WebUI, select ixl-v2.0.vae.safetensors from the VAE dropdown menu in Settings -&gt; Stable Diffusion.</li><li>Alternatively, most UIs allow you to select the VAE alongside the main checkpoint.</li></ol>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.

Illustrious-XL-v2.0

This is the first public release of the VAE, extracted directly from the Illustrious XL v2.0 Base model.We are releasing it as a standalone component for the community to use with other SDXL models. It encapsulates the core visual DNA of the Illustrious XL 2.0 project.Key characteristics of this VAE:<ul><li>Source: Extracted from Illustrious XL v2.0 Base. This ensures it's highly versatile and not over-stylized, making it an excellent general-purpose VAE.</li><li>Performance: Optimized for stable and artifact-free rendering across a wide range of resolutions (512px to 1536px).</li><li>Aesthetics: Delivers vibrant colors, high contrast, and the clean visual output that our Illustrious XL series is known for.</li><li>Compatibility: Recommended for use with any SDXL 1.0-based checkpoint.</li></ul>We recommend using this VAE to enhance the visual quality and color depth of your generations, especially if you've noticed issues with desaturation or blurriness with the default SDXL VAE.Enjoy

illustriousXLV20_v10.safetensors

v1.0

Just wanted to upload my version of the beautiful Finana Ryugu

Finana Ryugu

danbooruE621AutocompleteTag_22122024Refined.zip

v2.0

I just combined two of my favorite models, and got a pretty good model that works great with loras. Nochekaiser881's lora is good for this. I use for nswf generation.Sampling method DPM++ 2M SDESchedule type AutomaticHires NoneWidth 1024Height 1024CFG Scale 6 Sampling steps 20-40

AidenPear's

illustriousXL_v20.safetensors

aidenpears_v10.safetensors

smiling, looking at viewer, flexing

Nick

Better_handsome_male_face.safetensors

Handsome_Male_models_v2.safetensors

Handsome_young_male_models_face.safetensors

bananaSplitzXLNEWXXL_bananaSplitzXXLV11.safetensors

Ogawa Shizuka/Sakura Rio's 2000's Art Style - SDXL

checkpoint-e18_s684.safetensors

<h3 id="noobai-xl-vae-the-official-vae-for-the-epsilon-series-jw5bka30o">NoobAI XL VAE - The Official VAE for the Epsilon Series</h3>This is the official Variational Autoencoder (VAE) extracted from the popular NoobAI XL Epsilon Prediction v1.1 (noobaiXLNAIXL_epsilonPred11Version). It is the key to unlocking the signature visual style of NoobAI XL on any SDXL-compatible model.If you love NoobAI's gritty, high-contrast, and dramatically colored illustrative aesthetic, this VAE is an essential tool for your collection.Key Features<ul><li>Signature NoobAI Aesthetic: This VAE is fine-tuned to reproduce the core visual identity of NoobAI: punchy contrast, deep and dramatic shadows, and the ability to handle extreme, vivid color palettes without losing detail.</li><li>Stylized &amp; Illustrative Finish: Perfect for creators aiming for a non-photorealistic, "artsy" look. It enhances textures and adds a painterly, slightly gritty quality that makes images pop, moving them away from a generic, soft-focus feel.</li><li>Reduces "Muddy" Outputs: Specifically addresses the common issue of colors appearing washed out or blacks looking gray. This VAE ensures that dark themes are truly dark and that vibrant colors maintain their saturation and intensity.</li><li>Broad Compatibility: Extracted from the user-friendly Epsilon Prediction version of NoobAI, this VAE is designed for maximum compatibility with the vast majority of SDXL 1.0 models. It's a simple way to add a dose of NoobAI's style to your favorite checkpoints.</li></ul>How to Use<ol><li>Place this file in your stable-diffusion-webui/models/VAE folder.</li><li>In your UI, select this VAE from the VAE dropdown menu (often in Settings -&gt; Stable Diffusion).</li><li>Generate as usual and enjoy the enhanced contrast and color!</li></ol>Note: This VAE will not change the content or composition of your generation. It works during the final decoding step to refine colors, lighting, and fine details, giving your images the distinctive NoobAI XL finish.

NoobAI-XL (NAI-XL)

This is the first public, standalone release of the VAE from the NoobAI XL Epsilon Prediction v1.1 model.We are releasing it to allow the community to apply the unique and sought-after NoobAI visual style to a wider range of models.Key characteristics of this VAE:<ul><li>Source: Extracted directly from noobaiXLNAIXL_epsilonPred11Version.safetensors.</li><li>Purpose: This version of the main model was the result of additional training epochs specifically aimed at "improving the production rate of high-quality images and the representation of extreme colors." This VAE carries that exact DNA.</li><li>Performance: Excels at creating high-contrast scenes with rich, saturated colors and deep, true blacks. It is the perfect choice for dark themes, stylized characters, and any image that needs a strong visual impact.</li><li>Compatibility: As it comes from a standard Epsilon Prediction model, it is highly compatible and recommended for use across the SDXL ecosystem.</li></ul>We hope you enjoy bringing the powerful NoobAI aesthetic to all your favorite models

Files with SHA256: d7ed5ca5fd33ff846cf65f2efbb36d7a259258cbb70da71e711b2c612694ac00