New ComfyUI Node for Controlling the Krea2 Model

🛠 New ComfyUI Node for Controlling the Krea2 Model

ComfyUI-Krea2TextEncoder has been released, which uses vision-aware text conditioning via the Qwen3-VL-4B model. This allows for adding reference images and masks for precise prompt refinement.

🌍 Expands control capabilities in multimodal DiT models, allowing visual references to be used as part of the text prompt through a VLM encoder.

👤 Enables more precise control over style and details when generating in Krea2 by using specific areas on sample images.

Source 1: https://github.com/ethanfel/ComfyUI-Krea2TextEncoder