extract-prompts

a CLI tool to extract ComfyUI workflow JSON and A1111 prompts from images and videos.

Installation

npm install -g @craftgear/extract-prompts

Usage

extract-prompts <files...> [options]

Examples

# Extract from a single image
extract-prompts image.png

# Extract from multiple files with glob patterns
extract-prompts *.png *.webm

# Pretty format output
extract-prompts *.png --pretty

# Save extracted workflows to directory
extract-prompts *.png --save ./workflows

# Save to directory with spaces (use quotes)
extract-prompts *.png --save "output directory with spaces"

# Save to input file directory (uses directory of first input file)
extract-prompts *.png --save

# Convert A1111 parameters to ComfyUI workflow format
extract-prompts a1111_image.png --convert-a1111

# Combine conversion with pretty output
extract-prompts a1111_image.png --convert-a1111 --pretty

# Quiet mode (suppress non-error output)
extract-prompts *.png --quiet

Options

• -p, --pretty: Human-readable output format (default: JSON)
• -s, --save [directory]: Save workflows to directory (defaults to input directory if not specified)
• -q, --quiet: Suppress non-error output
• --overwrite: Overwrite existing files when saving
• --name-pattern <pattern>: File naming pattern (source|sequential|timestamp) [default: source]
• --organize <mode>: Organize saved files (none|format|date) [default: none]
• --convert-a1111: Convert A1111 parameters to ComfyUI workflow format

Supported Formats

Images

• PNG
• JPEG/JPG
• WebP

Videos

• MP4
• WebM
• MOV

Output Formats

JSON (default)

Raw JSON output with full workflow data.

Pretty (--pretty flag)

Human-readable format showing:

• LoRA models and strengths
• Prompts (positive/negative when detected)
• Sampler settings (steps, CFG, scheduler, seed)
• Model information
• Workflow statistics

Features

• ComfyUI Workflow Extraction: Extracts complete workflow JSON from ComfyUI-generated content
• A1111 to ComfyUI Conversion: Convert A1111 parameters to ComfyUI workflow format with full LoRA and upscaler support
• Prompt Detection: Intelligently identifies and extracts prompts from various node types
• LoRA Support: Extracts LoRA models and their strengths from multiple loader formats
• Upscaler Support: Handles hires.fix and upscaler parameters in conversions
• Conservative Labeling: Only shows positive/negative distinction when confident
• Batch Processing: Process multiple files at once with glob patterns
• Flexible Output: Multiple output formats for different use cases

Supported Node Types

• CLIPTextEncode
• WanVideoTextEncode
• Text Multiline
• easy showAnything
• KSampler / WanVideoSampler
• Power Lora Loader (rgthree)
• LoraTagLoader
• Various model loaders

Examples

Basic extraction

extract-prompts generated_image.png

Pretty format with LoRA and prompt details

extract-prompts workflow.webm --pretty

Save workflows organized by date

extract-prompts *.png --save ./extracted --organize date

Save to directory with spaces

# Use quotes for directory paths with spaces
extract-prompts *.png --save "My Output Directory"
extract-prompts *.png --save 'Output Folder With Spaces'

A1111 to ComfyUI Conversion

# Convert A1111 parameters to ComfyUI workflow
extract-prompts a1111_image.png --convert-a1111

# Example with LoRA tags in A1111 prompt:
# "beautiful girl <lora:style1:0.8> <lora:character:0.6>, masterpiece"
# Will generate ComfyUI workflow with LoRA loader nodes

# Example with upscaler parameters:
# Hires.fix: true, Hires upscaler: ESRGAN_4x, Hires steps: 10, Hires denoising: 0.5
# Will generate ComfyUI workflow with upscaler nodes for 2-pass generation

License

MIT