CyberRealistic Krea 2 - ComfyUI Workflow — v1.0 — v3.5 (Prompt Presets)

This ComfyUI workflow generates Krea 2 images from text prompts and can use an input image for image-to-prompt work; it includes previews and image comparisons.

With Vision disabled, the preset menu shows prompt-rewriting presets; enabling Vision switches it to image-to-prompt presets.

The workflow also offers optional Face Detailer and SesquiLSR latent upscaling, plus LoRA and NegPiP prompting.

PreviewImage and Image Comparer nodes show previews and comparisons, and a metadata node saves image metadata.

Connect a local LLM for prompt rewriting or image-to-prompt work; passthrough prompting works without LM Studio or another local LLM.

Listed ComfyUI node packs: comfyui-cyberdelia-metadata, comfyui-cyberdelia-z-engineer, comfyui-impact-pack, comfyui-impact-subpack, comfyui-krea2t-enhancer, comfyui-negpip-zimage, cyberkrea-sampler, and rgthree-comfy.

Listed model files: bbox/face_yolov8m.pt, CyberRealistic_Krea2_V3.0_BF16.safetensors, qwen_image_vae.safetensors, qwen3vl_4b_bf16.safetensors, and sam_vit_b_01ec64.pth.

In Cyberdelia Prompt Controls, enter a prompt, choose engineered (LLM) or passthrough (RAW), toggle Vision, and select a system-prompt preset.

This is release v3.5 (Prompt Presets), based on Krea 2.

No download URLs are listed for CyberRealistic_Krea2_V3.0_BF16.safetensors, qwen_image_vae.safetensors, qwen3vl_4b_bf16.safetensors, or sam_vit_b_01ec64.pth; locate these files separately before setup.

Sugestão