In simple English
This tool lets you run the Krea 2 image generation model on your own hardware. You can turn text prompts into high-resolution pictures using either a base model or a fast turbo model. The base model works best for training custom styles, while the turbo model creates pictures in just a few steps. You need a powerful graphics card with plenty of memory to run it. The software is free to download under an Apache license, but the model weights have special terms for commercial use.
What it is
It is the official Python repository for Krea 2, providing command-line scripts to generate images from text prompts using local model weights. The package splits into two distinct checkpoints for training and fast inference.
What it is good for
Solo builders who want to generate custom images locally or train personal style adapters without relying on paid cloud services.
- Includes both a raw base model for training and a fast turbo model for quick generation.
- Generates pictures up to 2K resolution with adjustable settings for guidance and steps.
- Provides clear instructions for running inference and integrating with popular tools like ComfyUI.
- Requires a strong graphics card with significant GPU memory to run local inference.
- Model weights use a custom community licence with restrictions on commercial use without a paid agreement.
- Setup requires downloading large checkpoint files from Hugging Face before you can start.
Paid cloud image generation services and closed-source art platforms.
ComfyUI for node-based workflows and Hugging Face Diffusers for custom training.
- Comfy-Org/ComfyUI141k stars
A node-based graphical interface that also supports running Krea 2 models locally.
- huggingface/diffusers35k stars
A standard Python library for loading and fine-tuning diffusion models like Krea 2.
- ggml-org/llama.cpp132k stars
A C and C++ library and command-line tool for running language models locally.
- EleutherAI/lm-evaluation-harness14k stars
Runs standard question sets against a language model and reports the scores.
- mudler/LocalAI50k stars
Serves text, vision, speech, image and video models from one self-hosted service that imitates the paid interfaces.
uv sync export OSS_RAW=/path/to/raw/safetensors export OSS_TURBO=/path/to/turbo/safetensors uv run inference.py "a fox walking in the snow" --checkpoint oss_turbo --steps 8 --cfg 0.0 --mu 1.15 --width 2048 --height 2048