Tiled Diffusion & VAE extension for sd-webui

This extension is licensed under CC BY-NC-SA, everyone is FREE of charge to access, use, modify and redistribute with the same license.
You cannot use versions after AOE 2023.3.28 for commercial sales (only refers to code of this repo, the derived artworks are NOT restricted).

由于部分无良商家销售WebUI，捆绑本插件做卖点收取智商税，本仓库的许可证已修改为 CC BY-NC-SA，任何人都可以自由获取、使用、修改、以相同协议重分发本插件。
自许可证修改之日(AOE 2023.3.28)起，之后的版本禁止用于商业贩售 (不可贩售本仓库代码，但衍生的艺术创作内容物不受此限制)。

If you like the project, please give me a star! ⭐

The extension helps you to generate or upscale large images (≥2K) with limited VRAM (≤6GB) via the following techniques:

• Reproduced SOTA Tiled Diffusion methods
  • Mixture of Diffusers
  • MultiDiffusion
  • Demofusion
• Our original Tiled VAE method
• My original Tiled Noise Inversion method

Features

• Core
  • [x] Tiled VAE
  • [x] Tiled Diffusion: txt2img generation for ultra-large image
  • [x] Tiled Diffusion: img2img upscaling for image detail enhancement
  • [x] Regional Prompt Control
  • [x] Tiled Noise Inversion
• Advanced
  • [x] [ControlNet support]()
  • [x] StableSR support
  • [x] SDXL support
  • [x] [Demofusion support]()

👉 在 wiki 页面查看详细的文档和样例，以及由 @PotatoBananaApple 制作的 快速入门教程 👉 Find detailed documentation & examples at our wiki, and quickstart Tutorial by @PotatoBananaApple 🎉

Examples

⚪ Txt2img: generating ultra-large images

prompt: masterpiece, best quality, highres, city skyline, night.

⚪ Img2img: upcaling for detail enhancement

original	x4 upscale

⚪ Regional Prompt Control

region setting	output1	output2

⚪ ControlNet support

original	with canny

	重绘 “清明上河图”
original
processed

⚪ DemoFusion support

original	x3 upscale

License

Great thanks to all the contributors! 🎉🎉🎉
This work is licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.