1. Workflow Overview

This "Everything is Designable" workflow specializes in artistic typography and background synthesis, powered by Qilin-revAnimated model and Floral Typography LoRA. Ideal for posters/logo/wallpaper design.

2. Core Models

• Base Model: Qilin-revAnimated_v122 (anime-optimized)

• LoRA: Floral Typography v1.0 (pattern/lettering focus)

• ControlNet: Tile model for detail enhancement

3. Key Nodes

Dependencies:

• Example image 12.png (ControlNet guide)

• Negative prompt embedding bad_prompt_version2-neg.pt

4. Pipeline Stages

Stage 1: Model Loading

• Load Qilin model + Floral LoRA (0.8 strength)

• Dual CLIP encoders for prompt processing

Stage 2: Conditioning

• Tile ControlNet at 0.75 strength

• Reference image affects composition

Stage 3: Generation

• Fixed seed 75663518246296 + Euler sampler

• Default 512x512 output (scalable to 1024x1024)

5. I/O Specification

• Inputs:

  • Positive prompt (must start with hua,)

  • Optional reference image

• Output:

  • Design images (saved to ComfyUI folder)

6. Critical Notes

⚠️ Requirements:

• Install easy-use extension for resolution node

• Download negative prompt embeddings separately

🔧 Pro Tips:

• Change batch_size to 4 in EmptyLatentImage for parallel generation

• Replace gradient background with other style keywords