1. Workflow Overview

This workflow generates cyberpunk-style art covers with integrated text rendering, combining image generation, captioning, upscaling, and typography effects. Key features:

• Reverse-engineers image descriptions using Meta-Llama-3.

• Generates cyberpunk images (Stable Diffusion + Flux LoRAs).

• Adds artistic fonts (e.g., "splash brush letters").

• Enhances resolution via UltimateSDUpscale (1024x1024).

Core Models:

• Stable Diffusion (F.1 Base): Main generation model.

• Meta-Llama-3.1-8B: Image captioning.

• R-ESRGAN_4x+ Anime6B: Upscaling.

• Flux LoRAs: Neon/mechanical style enhancements.

2. Key Components

Critical Nodes:

1. Joy_caption / Joy_caption_two:

   • Uses Meta-Llama-3 to generate image descriptions.

   • Install via ComfyUI Manager (unsloth/Meta-Llama-3.1-8B).

2. UltimateSDUpscale:

   • Upscales images with tile-based processing.

   • Requires R-ESRGAN_4x+ Anime6B in models/upscale_models.

3. LoraLoader:

   • Loads style LoRAs like:

     • Flux.1_Cyber-Electronic: Mechanical textures.

     • AsurFont02-F.1: Calligraphy font effects.

4. ModelSamplingFlux:

   • Optimizes SD sampling for finer details.

Dependencies:

• Flux models (download from LibLibAI to models/loras).

3. Workflow Structure

Three Groups:

1. Image Captioning (Group 35):

   • Input: Source image (e.g., output (26).png).

   • Output: Bilingual text description.

2. Image Generation (Group 34):

   • Input: Prompt, seed (e.g., 48473066447724), resolution (1024x1536).

   • Output: Initial cyberpunk image with fonts.

3. Upscaling (Group 33):

   • Input: Low-res image.

   • Output: 1024x1024 HD image (preserves font details).

4. Input & Output

Input Parameters:

• Image: Source file (PNG/JPG).

• Prompt: E.g., "Angel statue with neon lights, splash brush letters."

• Seed: Fixed value for reproducibility.

• Resolution: Customizable (e.g., 1024x1536).

Output:

• Final image (PNG) with cyberpunk style and typography.

• Intermediate results can be compared via Image Comparer.

5. Notes

• VRAM: ≥12GB recommended (UltimateSDUpscale tile size: 32x32).

• Compatibility:

  • Flux LoRAs require F.1 base model.

  • Install Impact Pack and rgthree extensions if nodes are missing.

• Fonts: Adjust AsurFont02-F.1 weight (default: 1.2) for intensity.