xiangyuT-ComfyUI-PromptEnhance/examples

Example Workflows

This directory contains example workflows demonstrating the usage of ComfyUI-PromptEnhance nodes.

Available Workflows

1. SDXL Prompt Enhance Workflow (sdxl_prompt_enhance_workflow.json)

A complete SDXL workflow that demonstrates how to use the Prompt Enhance Advanced node to enhance prompts before generating images.

Features:

• Uses the PromptEnhanceAdvanced node to enhance simple prompts into detailed visual descriptions
• Displays both the enhanced and original prompts for comparison using ShowText nodes
• Complete SDXL image generation pipeline
• Generates 1024x1024 images using SDXL base model

Workflow Components:

1. Prompt Enhancement Section:

   • Original Prompt Input: A primitive string node where you enter your simple prompt
   • Prompt Enhance Advanced (LLM): The main enhancement node that calls an LLM API to enhance your prompt
   • Enhanced Prompt Preview: Shows the enhanced prompt generated by the LLM (for debugging/demonstration)
   • Original Prompt Reference: Shows the original prompt for comparison

2. Image Generation Section (SDXL):

   • CheckpointLoaderSimple: Loads the SDXL base model (sd_xl_base_1.0.safetensors)
   • CLIPTextEncode (Positive): Encodes the enhanced prompt for image generation
   • CLIPTextEncode (Negative): Encodes negative prompt to avoid unwanted features
   • EmptyLatentImage: Creates a 1024x1024 latent image
   • KSampler: Generates the image using the enhanced prompt
   • VAEDecode: Decodes the latent image to pixels
   • SaveImage: Saves the generated image

How to Use:

1. Prerequisites:

   • Ensure you have SDXL base model installed in your ComfyUI models directory
   • Get an API key from DeepSeek (or another OpenAI-compatible API provider)

2. Load the Workflow:

   • In ComfyUI, click "Load" and select sdxl_prompt_enhance_workflow.json

3. Configure the Prompt Enhance Node:

   • Locate the "Prompt Enhance Advanced (LLM)" node
   • Replace your-api-key-here with your actual API key
   • (Optional) Modify the api_endpoint if using a different provider
   • (Optional) Modify the model name if using a different model

4. Enter Your Prompt:

   • In the "Original Prompt Input" node, enter your simple prompt
   • Example: "a beautiful sunset over mountains"

5. Generate:

   • Click "Queue Prompt" to run the workflow
   • The enhanced prompt will be displayed in the "Enhanced Prompt Preview" node
   • The generated image will be saved automatically

Expected Output:

• You will see the enhanced prompt in the ShowText node, which transforms simple prompts into detailed, visual descriptions
• The generated image will use the enhanced prompt for better quality and detail
• You can compare the original and enhanced prompts side-by-side

Customization:

• Adjust temperature in the Prompt Enhance node to control creativity (lower = more focused, higher = more creative)
• Modify max_tokens to control the length of the enhanced prompt
• Change the prompt_template to customize how prompts are enhanced
• Adjust sampler settings (steps, CFG scale, etc.) to control image generation quality

Notes

• The Prompt Enhance nodes require an active API key and internet connection
• API calls may take a few seconds depending on the provider and network speed
• The default template is written in Chinese and optimized for Chinese-language prompts, but it can also work with English prompts as the LLM will understand and respond appropriately. You may want to customize the prompt_template parameter for English-specific instructions if desired.
• The ShowText nodes are essential for debugging and understanding how your prompts are being enhanced

Troubleshooting

• API Key Error: Make sure you've entered a valid API key in the Prompt Enhance node
• Timeout: Increase the timeout parameter if API calls are timing out
• Model Not Found: Ensure the SDXL base model is downloaded and placed in the correct directory
• No Output: Check the console for error messages from the Prompt Enhance node