Image Generation AI with ComfyUI: Outpaint Edition

Progress 11 / 15
Table of Contents

Outpaint of Stable Diffusion can draw adding new content outside of original image, and extend original image naturally. For example, in state where part of photo or painting is cut, automatically complement surrounding of that cut part, and can show as if it existed from beginning.

In this article, I explain specific procedure to perform Outpaint with StableDiffusion using ComfyUI.

Work Flow

1. Introduction of ComfyUI

Please refer to this article.

【Stable Diffusion】ComfyUIを使って画像生成AIで遊んでみよう【導入編】

>-

blog.otama-playground.com

2. Creation of Workflow

Load default workflow, add yellow node like image below and connect.

Workflow
Workflow
Expansion
Expansion

3. Decide size and shape of image

Decide size of image to generate by tweaking parameters with Upscale Image node and Pad Image for Outpainting node.

Node to tweak size
Node to tweak size

4. Generate

Write what you want to generate in Outpaint part in prompt and generate.

If boundary with extended part appears clearly, increase value of Feathering, or decrease cfg.

Note: Feathering is parameter how much to blur boundary of extended part and original image.

Generation Result

Mask Image
Mask Image
Outpaint Result
Outpaint Result

I feel image quality became slightly bad but I could Outpaint seemingly well.

Conclusion

I tried using it, but defect is that quality of originally existing part becomes bad.

If you want to try other techniques related to image generation, please utilize the link collection below.

Stable Diffusionガイド:画像生成に役立つリンク集

>-

blog.otama-playground.com