Analysis updated 2026-05-18
Extend an existing image beyond its edges using AI-generated content that matches the original.
Build a ComfyUI workflow that outpaints images with the Krea 2 Turbo model.
Restore exact original pixels after generation with a smooth blended seam.
Experiment with reference-based image conditioning inside a node-based AI pipeline.
| realrebelai/comfyui-rebels-krea2-outpaint | 0petru/sentimo | agi-eval-official/forte | |
|---|---|---|---|
| Stars | 17 | 17 | 17 |
| Language | Python | Python | Python |
| Setup difficulty | hard | moderate | hard |
| Complexity | 4/5 | 3/5 | 4/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires a ComfyUI build with Krea 2 support plus manually downloading the matching model, VAE, CLIP, and LoRA files.
This project is a set of custom nodes for ComfyUI, a visual tool for building AI image generation workflows, that adds support for outpainting with a specific AI model called Krea 2 Turbo. Outpainting means extending an existing image beyond its original edges, generating new image content that flows naturally from what was already there. This pack ports a technique originally built for a different AI framework so it can run inside ComfyUI's node-based interface. The core idea is that the original image is placed onto a larger canvas at an exact position, so the AI model knows precisely which pixels are already known and which area needs to be generated. After the new content is generated, the pack restores the exact original pixels in their place and blends the seam so the transition looks smooth. The README is written for people already experienced with ComfyUI and AI image generation pipelines, and it assumes familiarity with concepts like models, LoRAs, which are small add-on files that adjust how a base model behaves, and node-based visual workflows. To use it, someone needs a ComfyUI setup that supports the Krea 2 model, along with the matching diffusion model, image decoder, and text encoder files, plus a specific LoRA file that must be downloaded separately and placed in the right folder. The pack adds four nodes that handle setting up the canvas and placement, encoding the prompt with a reference to the known image area, patching the AI model to use this specific outpainting technique, and compositing the final result. The README also documents recommended generation settings and explains which directions an image can currently be extended in, noting that this version only supports extending in one direction at a time rather than filling in a fully enclosed empty area. The node code in this pack is released under the MIT license, while parts of the underlying technique are adapted from other projects under MIT and Apache-2.0 licenses. The AI model weights themselves follow a separate Krea 2 Community License.
ComfyUI custom nodes that add AI image outpainting support for the Krea 2 Turbo model.
Mainly Python. The stack also includes Python, ComfyUI, PyTorch.
The node code is MIT licensed, free to use and modify, the AI model weights follow a separate Krea 2 Community License with restrictions on commercial misuse.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.