Krea2 workflow | ti2i 16GB VRAM CivitAI metadata, upscaling, controlnet, FaceDetailer (+indexing elements), inpainting ComfyUI beginner friendly K2 — v2 — v7

This Krea 2 ComfyUI workflow is for creating and editing images from text prompts or a loaded image. It includes inpainting, depth ControlNet, face detailing, and upscaling, with images available to preview and save.

Ollama prompt generation is included, with nodes for checking connectivity, generating prompts, and setting options.

LoadImage and CLIPTextEncode nodes provide image and text inputs.

PreviewImage and Image Saver nodes provide image preview and saving.

For ControlNet, the setup instructions call for cloning github.com/facok/comfyui-krea2-controlnet into custom_nodes and obtaining depth-control-lora.safetensors.

The linked Ollama model file is gemma-4-12B-it-uncensored-heretic.Q8_0.gguf.

Listed ComfyUI node packs: ComfyUI_Comfyroll_CustomNodes, comfyui_controlnet_aux, comfyui-custom-scripts, comfyui-image-saver, comfyui-impact-pack, comfyui-inpaint-cropandstitch, and rgthree-comfy.

Listed model files: bbox/face_yolov8m.pt, depth-control-lora.safetensors, ema_vae_fp16.safetensors, krea2_raw-Q6_K.gguf, Krea2_TextFusion_Refusal_Reduction.safetensors, qwen3-vl-4b-heretic_int8_dynamic_convrot.safetensors, seedvr2_ema_7b-Q4_K_M.gguf, and Wan2_1_VAE_fp32.safetens[

This is version v7, with Krea 2 as its base model.

The v7 download is the Config file blinksKrea2SimplePhoto_v7.json; it does not include model weights.

Before installing, verify which comfyui_controlnet_aux package is intended; the name alone does not identify a specific package.

Vorschlag · nicht geprüft