MultiDiffusion Upscaler – Tiled Diffusion & VAE Extension for Stable Diffusion WebUI

MultiDiffusion Upscaler – Tiled Diffusion & VAE Extension for Stable Diffusion WebUI

Category: Computer Vision
License: Creative Commons
Model Type: Image Generation
The MultiDiffusion Upscaler is an open-source extension for the Automatic1111 Stable Diffusion WebUI. It enables the generation and upscaling of ultra-high-resolution images (2K–16K+) on systems with limited VRAM (as low as 4–6GB) by employing advanced tiling techniques. By dividing images into manageable tiles and processing them individually, the extension allows for seamless and efficient image synthesis and enhancement without compromising quality.

Key Features

  • Tiled Diffusion: Facilitates text-to-image generation and image-to-image upscaling for ultra-large images.
  • Tiled VAE: Optimizes Variational Autoencoder processing to handle large images efficiently.
  • Tiled Noise Inversion: Introduces a novel method for improved upscaling quality.
  • Regional Prompt Control: Allows for different prompts in specified regions of an image.
  • ControlNet Support: Integrates with ControlNet for enhanced control over image generation.
  • StableSR Support: Compatible with StableSR for super-resolution tasks.
  • SDXL Support: Supports Stable Diffusion XL models for higher fidelity outputs.
  • Demofusion Support: Incorporates Demofusion techniques for advanced image synthesis.
GitHub

Project Screenshots

Project Screenshot
Project Screenshot
Project Screenshot

Similar Projects

Diffusers: State-of-the-Art Diffusion Models for Image, Video, and Audio Generation

Diffusers: State-of-the-Art Diffusion Models for Image, Video, and Audio Generation

Computer Vision
Contrastive Unpaired Translation (CUT)

Contrastive Unpaired Translation (CUT)

Computer Vision
Awesome Image Translation: Curated Resources on Image-to-Image Translation

Awesome Image Translation: Curated Resources on Image-to-Image Translation

Computer Vision
Anime2Sketch: A Sketch Extractor for Anime and Illustration

Anime2Sketch: A Sketch Extractor for Anime and Illustration

Computer Vision
Stable Diffusion: High-Resolution Text-to-Image Synthesis with Latent Diffusion Models

Stable Diffusion: High-Resolution Text-to-Image Synthesis with Latent Diffusion Models

Computer Vision
ruDALL-E: Russian Text-to-Image Generation

ruDALL-E: Russian Text-to-Image Generation

Computer Vision