Unlock Stunning Visuals: A Step-by-Step Guide to Stable Diffusion Workflow

📝 Workflow Overview

This workflow is designed to generate high-quality images with particle and light effects using Stable Diffusion.
It incorporates FluxGuidance for style enhancement and high-resolution upscaling to ensure final image clarity.

🧠 Core Models

1️⃣ Stable Diffusion (UNet)

• Function: The main image generation model responsible for producing images based on text prompts and control inputs.

• Model Used: F.1_Depth-fp16_1.0

• Installation:

  • Install via ComfyUI Manager.

  • Or manually download .safetensors and place it in models/checkpoints.

2️⃣ FluxGuidance

• Function: Enhances specific styles, making particle and light effects more pronounced.

• Parameter Used: guidance_strength = 2.5

• Installation:

  • Requires the Flux plugin, available through ComfyUI Manager.

3️⃣ CLIP (Text Encoder)

• Function: Converts text prompts into usable vectors for image generation.

• Model Used: clip_l, t5xxl_fp16

• Installation:

  • Install via ComfyUI Manager.

  • Or manually place it in models/clip.

4️⃣ VAE (Variational Autoencoder)

• Function: Enhances image quality by improving color and detail representation.

• Model Used: ae.sft

• Installation:

  • Obtain via ComfyUI Manager.

  • Or manually place it in models/vae.

5️⃣ Super-Resolution Upscaling

• Function: Increases the resolution of the final output for enhanced clarity.

• Model Used: 4x_NMKD-Siax_200k

• Installation:

  • Download 4x_NMKD-Siax_200k.pth and place it in models/upscale.

📦 Key Components (Nodes)

📂 Major Workflow Groups

1️⃣ Text Encoding

• Function: Converts text descriptions into usable vectors for Stable Diffusion.

• Key Components:

  • DualCLIPLoader

  • CLIPTextEncode

  • ConditioningZeroOut

2️⃣ Flux Style Enhancement

• Function: Adjusts the Flux style to emphasize particle and light effects.

• Key Components:

  • FluxGuidance

  • guidance_strength = 2.5

3️⃣ Sampling & Generation

• Function: Generates images based on input conditions.

• Key Components:

  • KSampler

  • Sampling Parameters:

    • Euler

    • 6 sampling steps

    • CFG Scale = 6

    • Fixed Seed

4️⃣ High-Resolution Upscaling

• Function: Enhances the final image resolution for better quality.

• Key Components:

  • ImageUpscaleWithModel

  • 4x_NMKD-Siax_200k

🔢 Inputs & Outputs

📥 Main Inputs

• Text Prompt (describing key aspects of the particle and light effects)

• "Realistic portrait photography, realistic skin texture, white background. Best quality, high quality, HD 8K"

• Sampling Parameters:

  • Seed Value (Fixed or Random)

  • Sampling Method (Euler)

  • Sampling Steps (6)

  • CFG Scale (6)

📤 Main Outputs

• Final high-quality image with light and particle effects

• Upscaled high-resolution version

• Output suitable for special effects or artistic compositions

⚠️ Important Considerations

1. Hardware Requirements

   • Requires at least 8GB GPU (12GB+ recommended).

   • FluxGuidance and super-resolution upscaling are memory-intensive.

2. Flux Style Adjustments

   • Adjust guidance_strength to control the intensity of particle and light effects.

3. Sampling Parameter Optimization

   • Euler is faster, while DPM++ provides better quality.

   • CFG Scale above 7 may cause oversaturation; recommended value: 6–7.

4. Super-Resolution Upscaling

   • Using 4x_NMKD-Siax_200k produces high-quality upscaling but increases processing time.

Conclusion

This ComfyUI workflow enhances particle and light effects using FluxGuidance, generates high-resolution special effects images with Stable Diffusion, and applies super-resolution upscaling for improved clarity.
It is ideal for sci-fi effects, artistic creations, and cinematic visuals.

Let me know if you need further modifications! 🚀