Fronyax/KREA2_Edit_SNOFS_Merge

UPDATE!

Regarding the Krea 2 Identity Edit LoRA v2, I'm still not entirely satisfied with it. Overall, v1.2 produces better results out of the box without tweaking anything. It just works, while v2 is still pretty hit or miss.

I suspect v2 requires more fine-tuning of the settings and LoRA weight to get consistently good results with SNOSF, just like v1.2. I'll experiment with it more and revisit it in the future.

And one more thing, this model is still enjoying its indefinite vacation in Civitai jail. No ETA on its release, of course. Guess we'll just have to wait for them to grant it parole 😂

---

A merge of Krea 2, the Krea 2 Identity Edit v1.2 adaptation by conradlocke, and SNOFS v1.4 LoKr by Ashen. Both adaptations are baked directly into the model, so the separate LoRAs are not required.

Requirements

Requires the ComfyUI-Krea2Edit node pack. The edit LoRA was trained using dual conditioning (in-context VAE tokens + image-grounded Qwen3-VL encoding), which stock nodes don't provide.

The example workflow is attached as an optional file.

Note: If you're using the BF16, INT8, or INT4 version, replace the GGUF loader node with the native Load Diffusion Model node.

Required Nodes

• ComfyUI-Krea2Edit
• ComfyUI-GGUF (GGUF versions only)

Credits

• Krea 2: Original base model by Krea AI
• Krea 2 Identity Edit: Adaptation by conradlocke
• SNOFS v1.4 LoKr: By Ashen (Hugging Face)

---

This is an unofficial community modification and is not affiliated with or endorsed by Krea.ai.