Architecture
Explore materials, lighting, interiors, facades, and early design directions with an open image model built for aesthetic control.
Krea2OpenSource
The most aesthetic open-source image model available. Train LoRAs on RAW, then run them fast on Turbo.
Reddit AMA 10am PST
ComfyUI Livestream 3pm PST Hugging Face
SGLang 
FAL Comfy 
Nous Research 
Replicate Cloudflare
Together AI Google GCP
Amazon AWS
Hugging Face
SGLang FAL Comfy Nous Research Replicate Cloudflare
Together AI Google GCP
Amazon AWS
Our Open Source Partners
The most aesthetic open-source image model
Krea 2 Open-Source brings Krea 2, our aesthetic-first model family, into the open with open weights on Hugging Face. Use Turbo for fast, polished text-to-image and RAW when you need the malleable base checkpoint for LoRA training in ComfyUI and beyond.
Prompt
Male cowboy riding horse through the desert, close up shot
Use cases
Build with Krea 2
API
Generate with Krea 2 from your own product, workflow, or backend through the Krea API.
Licensing
One place for license details.
Krea 2 is available with community and enterprise licensing paths depending on how you plan to use and deploy the model. The licensing page has the current agreement, commercial use details, and guidance for teams that need custom terms.
Read the licensing page
Start building with Krea 2 Open-Source.
Explore the open-source release, follow model updates, and read the technical writeup behind Krea 2 RAW and Turbo.
Gallery
Created with Krea 2
Questions about Krea 2 Open-Source
Krea 2 Open-Source is the open-source release of Krea 2. It ships as two checkpoints: Krea 2 RAW for training and research, and Krea 2 Turbo for fast, high-quality text-to-image inference.
Krea 2 RAW is the undistilled base checkpoint. It is diverse, malleable, and intended for fine-tuning, post-training, and LoRA training. Krea 2 Turbo is an 8-step distilled checkpoint for fast inference.
Train LoRAs on Krea 2 RAW, then apply them on Krea 2 Turbo. LoRAs trained on RAW are designed to express strongly on Turbo, so you can train on the base and run inference fast.
Use RAW when you want a malleable base checkpoint for fine-tuning or research. For polished, fast text-to-image outputs, use Krea 2 Turbo.
Krea 2 Turbo is an 8-step distilled checkpoint built for fast, high-quality text-to-image generation.
Krea 2 Open-Source uses a Qwen Image VAE, a 12B dense DiT backbone, and a Qwen3-VL text encoder with multi-layer feature aggregation.
Krea 2 Open-Source is for creators, researchers, and teams building with open models
Hugging Face
SGLang FAL Comfy Nous Research Replicate Cloudflare
Together AI Google GCP
Amazon AWS
Hugging Face
SGLang FAL Comfy Nous Research Replicate Cloudflare
Together AI Google GCP
Amazon AWS
Want to bring Krea 2 into your company or research workflow?
Contact sales