Krea2 workflow | ti2i 16GB VRAM CivitAI metadata, upscaling, controlnet, FaceDetailer (+indexing elements), inpainting ComfyUI beginner friendly K2 — v2 — v3

Use this ComfyUI workflow to make and edit Krea 2 images from an image and text prompt, with Ollama prompt enhancement and image preview and saving.

The graph includes Krea 2 control, ControlNet, inpainting, face-detailing and SeedVR2 video-upscaling nodes.

The graph has image and text-prompt input nodes.

Image preview, comparison and saving nodes provide ways to review and save images.

For ControlNet, the description says to clone comfyui-krea2-controlnet into custom_nodes and get the Krea-2 depth ControlNet LoRA.

For Ollama prompt enhancement, the description names gemma-4-12B-it-uncensored-heretic.Q8_0.gguf.

The workflow lists these 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 include bbox/face_yolov8m.pt, depth-control-lora.safetensors, ema_vae_fp16.safetensors, krea2_turbo-Q4_K_M.gguf and krea2_turbo-Q6_K.gguf.

Other listed model files are qwen_image_vae.safetensors, qwen3-vl-4b-heretic_int8_dynamic_convrot.safetensors, qwen3vl_4b_fp8_scaled.safetensors and seedvr2_ema_7b-Q4_K_M.gguf.

The graph includes Ollama options, a sampler selector and Image Saver input parameters.

Before setup, confirm which comfyui_controlnet_aux package to install and where to get ema_vae_fp16.safetensors, krea2_turbo-Q4_K_M.gguf, krea2_turbo-Q6_K.gguf and seedvr2_ema_7b-Q4_K_M.gguf; no source links are listed for these items.

Suggestion · non vérifié