raxephion/stormforge-krea2-webui — explained in plain English
Analysis updated 2026-05-18
Run Krea 2 Turbo locally through a simple web page instead of raw ComfyUI.
Generate images on a lower VRAM graphics card using built in memory saving defaults.
Swap in new Krea 2 model or encoder files and refresh the interface to use them.
| raxephion/stormforge-krea2-webui | 0marildo/imago | 0xdfi/glm-5.2-1m-4x-dgx-spark | |
|---|---|---|---|
| Stars | 3 | 3 | 3 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | easy | hard |
| Complexity | 3/5 | 2/5 | 5/5 |
| Audience | vibe coder | general | researcher |
Figures from each repo's GitHub metadata at analysis time.
Requires manually placing Krea 2, text encoder, and VAE model files into specific ComfyUI folders.
StormForge Krea2 WebUI is a local, simplified interface for running the Krea 2 Turbo image generation model on your own computer, aimed especially at people with limited graphics memory. Instead of building its own generation engine, it runs a hidden copy of ComfyUI in the background to do the actual work, since ComfyUI already has built in support for Krea 2 and can load the smaller fp8 model format while offloading data to save memory. On top of that hidden ComfyUI process sits a lighter, easier Gradio web page that the person actually interacts with. Getting started means running an install.bat file, which sets up an isolated Python environment and installs PyTorch, ComfyUI, and the app's own requirements. After that, three model files need to be manually copied into specific ComfyUI folders: the Krea 2 Turbo model itself, a Qwen3-VL text encoder, and a Qwen-Image VAE. Once those files are in place, running run.bat starts ComfyUI quietly in low VRAM mode and then opens the simplified Gradio interface in a browser window. The default generation settings mirror what Krea 2 Turbo's creators recommend, and a memory saving decode option is turned on by default, which can be switched off if a computer has more graphics memory to spare. If image generation fails, the README points users to a separate console window running the ComfyUI backend, which shows the full error details. If the graphics card is not detected, the install script's CUDA settings may need adjusting to match the installed driver. Whenever model files are added or replaced, the interface needs a manual refresh, or the whole app needs restarting, to pick up the change. The project is explicitly marked as still under active development.
A simplified local web interface for running the Krea 2 Turbo image model, built on top of a hidden ComfyUI backend for low VRAM systems.
Mainly Python. The stack also includes Python, Gradio, ComfyUI.
No license information is given in the README, so usage rights are unclear.
Setup difficulty is rated moderate, with roughly 1h+ to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
double-check against the repo, no cap.