This workflow is a smart AI-powered background creator designed for SDXL, Pony, Illustrious, and NoobAI users.
It uses Florence-2’s image understanding and a large language model (LLM) to automatically transform an uploaded image or written scene idea into detailed background prompts — perfect for generating cinematic, atmospheric, or stylized environments.

No need to write complex prompts — the workflow automatically builds scene-focused tags that emphasize lighting, mood, architecture, and natural settings, while avoiding people or creatures.

⚙️ Main Features

• 🖼️ AI Scene Understanding (Florence-2)
  Reads your image and produces a highly descriptive caption that captures the environment, lighting, and tone.

• ✍️ Smart Background Tag Generator (LLM)
  Converts the raw caption into clean, structured tag prompts ready for SDXL or other diffusion models.

• 🌄 Character-Free Backgrounds
  Automatically removes humans or creatures from tags, keeping the focus entirely on scenery and atmosphere.

• 💡 Customizable Style Guidance
  You can modify the system textbox to guide the LLM — e.g., “cinematic lighting,” “dreamlike fantasy,” “urban nightscape,” etc.

• ⚡ Instant Preview
  Built-in ShowText nodes display both the caption and the final background tags for easy reference or reuse.

🧠 How It Works

1. Load an image or describe a scene

2. Florence2 generates a detailed caption

3. The LLM (NemoMix 12B) refines it into short, comma-separated tags

4. Output: professional-quality background prompts you can feed directly into SDXL or any T2I model

🧰 Required Custom Nodes

• comfyui-florence2

• comfyui_vlm_nodes

• ComfyUI-Chibi-Nodes

• was-node-suite-comfyui

• comfyui-custom-scripts

🎨 Perfect For

• Creating cinematic background prompts for SDXL or Flux

• Environment concept art generation

• AI video backdrops, game scenes, or storytelling visuals

• Artists who want quick, repeatable scene compositions without manual tagging