1. Overview

This workflow manages fine-grained prompt categorization for generating stylized images. Key features:

• Multi-Text Input: 5 CR Text nodes for prompts (style/composition/character/etc.).

• Dynamic Merging: Text List + Text List to Text nodes concatenate prompts.

• HDR Control: LayerFilter: HDREffects enhances dynamic range.

• Ink Painting Style: Combines Neta Art XL model + Chinese Ink LoRA.

Core Models:

• Main: Neta Art XL 二次元角色 （更新V2）_V2.0

• LoRA: 国画 | 水墨 | 山水画 SDXL_水墨山水 XL (weight=0.8)

• VAE: sd_xl_vae_1.0

2. Key Nodes

Dependencies:

• Plugins:

  • ComfyUI-Custom-Nodes: For CR Text nodes.

  • LayerFilter: GitHub.

• Models: Download from LibLib.art (links in Chinese section).

3. Workflow Groups

1. Input Group:

   • Style: impasto, chinese style

   • Composition: medium shot, Overhead shot

   • Character: 1girl, long hair, Hanfu

   • Background: moon, night

   • Quality: best quality, masterpiece

2. Processing:

   • Text Merge: Text List → Text List to Text → CLIPTextEncode.

   • Negative Prompt: Direct input to CLIPTextEncode.

3. Generation:

   • Sampling: Euler, 40 steps, CFG=8, 768x1344 resolution.

   • HDR: Adjusts brightness(0.5)/contrast(0.25)/saturation(0.1).

4. Inputs & Outputs

• Inputs:

  • Prompts: Fill in 5 CR Text nodes.

  • HDR Params: Tweak in LayerFilter: HDREffects.

  • Seed: Fixed or random via easy seed.

• Output: HDR-enhanced image saved to ComfyUI/output.

5. Notes

1. Plugin Compatibility:

   • Install all custom nodes to avoid Missing Node errors.

2. VRAM: ≥8GB GPU for 768x1344 (enable xformers).

3. LoRA Weight: Default 0.8; higher may overfit style.

4. HDR Settings: Keep values 0-1 to avoid overexposure.