Krea 2 Open-Source Models are now available in ComfyUI
Krea ships two checkpoints — Krea 2 RAW and Krea 2 Turbo.
- Models
- Krea 2
- Reading Time
- 2 min
Workflow Overview
Krea ships two checkpoints — Krea 2 RAW and Krea 2 Turbo.
Content type: Workflow
Primary intent: Download
Required Models
- Krea 2
Setup Notes
- Install the required models before opening the workflow template.
- Use the download button above to import the workflow JSON into ComfyUI.
You can operate Krea 2 OSS locally in ComfyUI today. These open-weight checkpoints—Krea 2 RAW and Krea 2 Turbo—integrate with standard diffusion model nodes, enabling local hardware training and generation. Krea 2 OSS demonstrates proficiency in creating diverse artistic results and stands as a highly advised image model for adoption. For direct insights from Krea regarding these checkpoints, consult the GitHub repository and the technical report.
Download Krea 2 Turbo Workflow
Two models, designed to work together
Krea 2 OSS features two specialized checkpoints.
Krea 2 RAW serves as the foundational model: an unrestrained pretrained checkpoint optimized for diverse outputs and fine-tuning, performing optimally without additional processing.
Krea 2 Turbo represents a streamlined, distilled variant tailored for high-quality image generation efficiency (8 steps), delivering enhanced quality at modest variation.
Deploy Krea 2 RAW for adaptable training, then transition to Turbo for expedited application execution.
Model Highlights
Speed. Turbo streamlines generation to 8 steps, allowing swift iterations.
Stylistic range. Krea 2 OSS encompasses a broad aesthetic repertoire, from lifelike portraits to stylized anime visuals.
Malleability. RAW's undistilled framework supports extensive customization, providing diverse guides for targeted artistic adjustments.
Prompting
Opt for detailed, extensive prompts to achieve premium outcomes; the model's training accommodates varying prompt lengths and leverages integrated Prompt Enhancer directives.
Architecture
Krea 2 OSS utilizes a 12B dense DiT structure, incorporating a Qwen Image VAE and a Qwen3-VL text encoder equipped with feature aggregation capabilities.
Get started
Upgrade ComfyUI to version 0.26.0.
Secure the workflow here or access it in the template library.
Observe workflow notes to download required models and position them accurately in the model folder.