GitHub radar
Krea 2 image model now installs through ComfyUI
The open Krea 2 image model now ships as one ComfyUI package with ready templates and styles — no more Mac-only apps required.
Krea 2 is an open text-to-image model made by Krea AI, a company known for its online image editor. It ships in two forms. Raw is the untrained base, meant for people who train their own styles on top of it. Turbo is the fast version, built for finished images without a long wait. Until now Krea 2 was mostly installed through Mac apps like Draw Things and DiffusionBee. Comfy-Org, the team behind the popular local image tool ComfyUI, packaged both versions together with ready templates and a set of style presets. Because of that, the model now installs on Windows and Linux too, on any GPU, not only on a Mac. Turbo produces an image in eight steps, while a model of this kind usually needs more than fifty.
Why a vibe-coder should care
For projects already built around ComfyUI, this is the simplest way to add a current image model without paying for a subscription or sending prompts to someone else's server. Krea AI calls it the most aesthetic open model available — that is their own claim, not something I can check myself, but almost ten million downloads in a month is a real sign the community picked it up. The catch is that you still need a dedicated GPU and a working ComfyUI setup. This is not a one-click mobile app.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the Krea 2 image generation model for me through ComfyUI: https://huggingface.co/Comfy-Org/Krea-2 — pick the lightweight version for my GPU, load the ready-made template, and show me how to generate a first image.
A GPU with 12 GB of memory or more — the lightweight version of the whole ComfyUI package takes around 12 GB.
Open on Hugging Face▌ More finds