nodes/MushroomFleet-ComfyUI-Variationator

Files

T

Drift Johnson 8e3669847a Add files via upload

2025-11-25 09:29:54 +00:00

15 KiB

Raw Permalink Blame History

ComfyUI Variationator - Prompt Orchestration Nodes

Advanced prompt manipulation system integrated into ComfyUI workflows. Control lighting, style, and variations with precision using 20 pre-defined modifiers and unlimited transformations.

🔗 GitHub Repository: https://github.com/MushroomFleet/ComfyUI-Variationator

📦 Installation

cd ComfyUI/custom_nodes
git clone https://github.com/MushroomFleet/ComfyUI-Variationator.git
# Restart ComfyUI

See INSTALLATION.md for detailed installation instructions and troubleshooting.

🎯 What This Does

Brings professional prompt engineering capabilities directly into your ComfyUI workflows:

20 Lighting Modifiers: Sunrise to moonlit night, dry and wet conditions
Style Transformations: Cyberpunk, noir, minimalist, and custom styles
Smart Variation: Natural language prompt modification
Batch Generation: Multiple variations from single input
Preset Collections: Optimized modifier sets for common use cases

📦 Available Nodes (9 Total)

All nodes appear under the "Prompt Orchestration" category.

1. Lighting Modifier (Single)

Purpose: Apply one lighting modifier to a prompt

Inputs:

text: Your base prompt
modifier_id: Number 1-20 (see modifier reference below)
preserve_base: Append vs replace (default: True)

Output: Modified prompt with lighting applied

Example Use:

Input: "Sports car on highway"
Modifier: 7 (golden hour)
Output: "Sports car on highway, golden hour, warm low-angle side-light, clear sky"

2. Batch Variation Generator

Purpose: Generate multiple variations at once

Inputs:

text: Base prompt
category: "all", "dry", "wet", or "custom"
custom_ids: Comma-separated IDs if using "custom" (e.g., "1,7,18")

Outputs: Up to 10 variation strings (var_1 through var_10)

Example Use: Connect each output to a different CLIP Text Encode for batch image generation.

3. Style Transform

Purpose: Apply artistic style to prompt

Inputs:

text: Base prompt
style: Choose from 11 built-in styles or "custom"
maintain_composition: Keep original composition (default: True)
custom_style: Your custom style description (if style="custom")

Output: Style-transformed prompt

Built-in Styles:

cyberpunk, noir, minimalist, pop_art_1960s
bauhaus, art_deco, impressionist, futuristic
vintage, brutalist

4. Smart Variation (NLP)

Purpose: Natural language prompt modification

Inputs:

text: Base prompt
intent: What you want (e.g., "make it golden hour", "add wet street")

Output: Intelligently modified prompt

Example Intents:

"make it golden hour"
"convert to wet night scene"
"add cyberpunk style"
"morning sunrise lighting"

5. Preset Collection

Purpose: Apply optimized modifier sets

Inputs:

text: Base prompt
preset: Choose from 5 presets

Outputs: 8 variations optimized for the use case

Available Presets:

key_moments: Dramatic lighting variety (6 variations)
commercial: Product photography optimized (4 variations)
artistic: Creative with styles (7 variations)
full_day_cycle: Complete day progression (16 variations)
wet_night: Dramatic wet conditions (4 variations)

6. Modifier Info

Purpose: Display information about a modifier

Inputs:

modifier_id: Number 1-20

Outputs:

description: Full modifier description
category: "dry" or "wet"
tags: Comma-separated tags

Use: Connect to a text display to see what modifiers do.

7. Prompt Combiner

Purpose: Merge multiple text inputs

Inputs:

text_1 through text_4: Text pieces to combine
separator: String to join with (default: ", ")

Output: Combined text

Example:

text_1: "Sports car"
text_2: "on mountain road"
text_3: "golden hour"
separator: ", "
Output: "Sports car, on mountain road, golden hour"

8. Random Modifier

Purpose: Apply random modifier from category

Inputs:

text: Base prompt
category: "all", "dry", or "wet"
seed: Random seed for reproducibility

Outputs:

modified_prompt: Result with random modifier
used_modifier_id: Which modifier was selected

Use: Create diverse variations with randomness.

9. Multi-Style Layer

Purpose: Apply multiple styles sequentially

Inputs:

text: Base prompt
style_1, style_2, style_3: Up to 3 styles

Output: Prompt with layered styles

Use: Create complex aesthetic combinations.

🌟 The 20 Lighting Modifiers

DRY Category (1-16): Clear Conditions

ID	Time/Condition	Description
1	Sunrise	car-friendly side-light, clear sky
2	Early Morning	subtle back-light along road axis, clear sky
3	Late Morning	high-sun hard light for crisp body lines
4	Midday	crisp hard sunlight, clear sky
5	Early Afternoon	gentle side-light, clear sky
6	Late Afternoon	long shadow geometry, clear sky
7	Golden Hour ⭐	warm low-angle side-light (most popular)
8	Sunset	soft rim-lit edges, clear sky
9	Twilight	residual glow, clear sky
10	Blue Hour	deep clear sky
11	Night	clear sky with clean highlights
12	Moonlit Night	clear sky
13	Early Morning	three-quarter back-light, clear sky
14	Late Afternoon	subtle back-light along road axis
15	Evening	car-friendly side-light, clear sky
16	Dawn	clean low-haze air, clear sky

WET Category (17-20): Dramatic Wet Conditions

ID	Condition	Description
17	Predawn Blue Hour	wet road, subtle back-light
18	Blue Hour Wet ⭐	wet road, car-friendly side-light (dramatic)
19	Twilight Wet	wet road, residual glow
20	Night Wet	wet street, warm interior lights glowing

🔧 Common Workflow Patterns

Pattern 1: Single Image with Lighting

[Text Input] → [Lighting Modifier] → [CLIP Text Encode] → [KSampler]

Set modifier_id to 7 for golden hour (most versatile).

Pattern 2: Batch Variations

[Text Input] → [Batch Variation Generator] → [Multiple CLIP Text Encode nodes]
                                           → [KSamplers for each variation]

Connect each var_output to a separate CLIP encoder.

Pattern 3: Style + Lighting Combo

[Text Input] → [Lighting Modifier] → [Style Transform] → [CLIP Text Encode]

Apply lighting first, then style for best results.

Pattern 4: Smart Creative Workflow

[Text Input] → [Smart Variation] → [CLIP Text Encode]

Use natural language like "make it cyberpunk golden hour".

Pattern 5: Preset Collection for Comparison

[Text Input] → [Preset Collection] → [8 separate render paths]

Generate 8 variations optimized for a specific use case.

Pattern 6: Random Exploration

[Text Input] → [Random Modifier] → [CLIP Text Encode]
[Seed Control]    ↓
                [Display Used ID]

Generate different lighting each run, track which modifier was used.

💡 Usage Tips

Best Practices

Start with Golden Hour (7): Most universally flattering
Wet modifiers (17-20) for dramatic urban/automotive scenes
Use presets for quick, professional results
Layer styles carefully: Apply lighting before styles
Batch testing: Use Batch Variation Generator for exploration

Modifier Selection Guide

For Products:

Use modifiers: 3, 4, 7, 11
Or use "commercial" preset

For Automotive:

Morning/evening: 1, 2, 7, 8
Dramatic: 18, 20 (wet conditions)

For Architecture:

Full day: 1-16 (all DRY)
Dramatic: 7, 8, 10, 18

For Portraits:

Flattering: 7, 8, 15
Dramatic: 10, 18

Performance Tips

Batch node outputs: Can connect to batch processing nodes
Seed control: Use fixed seeds with Random Modifier for reproducibility
Text preview: Use Show Text nodes to preview before rendering
Save presets: Save workflow variations for different modifier sets

📋 Example Workflows

Example 1: Golden Hour Car Photography

[Prompt: "luxury sports car mountain road"]
    ↓
[Lighting Modifier: modifier_id=7]
    ↓
[CLIP Text Encode]
    ↓
[KSampler]

Result: Car with warm golden hour lighting

Example 2: Cyberpunk Style Variation

[Prompt: "city street night scene"]
    ↓
[Lighting Modifier: modifier_id=11 (night)]
    ↓
[Style Transform: style=cyberpunk]
    ↓
[CLIP Text Encode]

Result: Night scene with cyberpunk aesthetic

Example 3: Multi-Variation Comparison

[Prompt: "product on pedestal"]
    ↓
[Preset Collection: preset=commercial]
    ↓
[var_1] → [CLIP] → [KSampler 1]
[var_2] → [CLIP] → [KSampler 2]
[var_3] → [CLIP] → [KSampler 3]
[var_4] → [CLIP] → [KSampler 4]

Result: 4 product shots with optimal lighting

Example 4: Smart Natural Language

[Prompt: "architectural exterior"]
    ↓
[Smart Variation: intent="make it golden hour with dramatic shadows"]
    ↓
[CLIP Text Encode]

Result: Automatically applies golden hour + emphasis on shadows

Example 5: Random Daily Inspiration

[Prompt: "creative scene"]
    ↓
[Random Modifier: category=all, seed=<daily seed>]
    ↓
[Modifier Info: <connect used_modifier_id>]
    ↓
[CLIP Text Encode]

Result: Different lighting each day, see which modifier was chosen

🔄 Integration with Other Nodes

With Wildcard Nodes

[Wildcard: "__subject__"] → [Lighting Modifier] → [CLIP]

Combine dynamic subjects with lighting control.

With Lora/Checkpoint Switching

[Text] → [Batch Variation] → [CLIP per variation]
                           → [Different LoRAs per output]

Test same prompt with different lighting and models.

With Conditioning Combine

[Base Prompt] → [Lighting Modifier] → [CLIP] ─┐
[Detail Prompt] → [CLIP] ────────────────────┬→ [Conditioning Combine]

Layer lighting-modified base with detailed conditioning.

🎨 Advanced Techniques

Technique 1: Progressive Lighting

Use multiple Lighting Modifier nodes in sequence to build complex lighting:

[Base] → [Modifier 7: golden hour] → [Modifier 19: wet road] → [Result]

Creates golden hour with wet reflections.

Technique 2: Style Layering

[Base] → [Style: noir] → [Style: minimalist] → [Multi-layered aesthetic]

Use Multi-Style Layer node or chain Style Transform nodes.

Technique 3: Conditional Lighting

                    ┌→ [Modifier 7] → [Day scene]
[Base] → [Switch] ──┤
                    └→ [Modifier 18] → [Night scene]

Use switch nodes to choose lighting based on conditions.

Technique 4: Batch A/B Testing

[Prompt] → [Batch Variation: custom_ids="7,18"]
    ↓
[var_1: golden hour] → [Render]
[var_2: blue hour wet] → [Render]

Compare specific modifiers side-by-side.

📊 Node Reference Quick Chart

Node	Input Type	Output Type	Best For
Lighting Modifier	Single text	Single text	Precise control
Batch Variation	Single text	10 texts	Exploration
Style Transform	Single text	Single text	Aesthetic change
Smart Variation	Single text	Single text	Quick natural edits
Preset Collection	Single text	8 texts	Professional sets
Modifier Info	Number	3 texts	Understanding modifiers
Prompt Combiner	4 texts	Single text	Building prompts
Random Modifier	Text + seed	Text + number	Randomization
Multi-Style	Text + styles	Single text	Complex aesthetics

🐛 Troubleshooting

Issue: Node outputs empty text

Cause: Invalid modifier ID or missing inputs
Solution: Check modifier_id is 1-20, ensure text input is connected

Issue: Batch outputs are all the same

Cause: Not enough modifiers in selected category
Solution: Use "all" category or specify more custom_ids

Issue: Style not applying

Cause: Style name typo or "custom" selected without custom_style input
Solution: Double-check style dropdown selection

Issue: Smart Variation not working as expected

Cause: Intent too vague or unrecognized keywords
Solution: Use specific terms like "golden hour", "wet", "night", "cyberpunk"

🎓 Learning Path

Start Simple: Try Lighting Modifier with modifier_id=7
Explore: Use Batch Variation with category="all"
Style: Add Style Transform node after Lighting Modifier
Optimize: Use Preset Collection for your use case
Advanced: Combine multiple nodes for complex effects

💾 Workflow JSON Examples

See the example_workflows/ folder for:

basic_lighting.json - Simple lighting modifier workflow
batch_comparison.json - Multi-variation comparison
style_transform.json - Style transformation workflow
commercial_product.json - Product photography setup
random_exploration.json - Random modifier with seed control

Core System Docs: ../README.md
Installation Guide: INSTALLATION.md
Node Reference: NODE_REFERENCE.md
Python API: ../prompt_orchestrator.py

📝 Version Info

Version: 1.0.0
Repository: https://github.com/MushroomFleet/ComfyUI-Variationator
ComfyUI Compatibility: Latest version
Python: 3.7+
Dependencies: None (stdlib only)

🚀 Quick Start Workflow

Create your first workflow in 3 steps:

Add nodes:
- Add Node → Prompt Orchestration → Lighting Modifier (Single)
- Connect your text input
Configure:
- Set modifier_id: 7 (golden hour)
- Keep preserve_base: True
Connect:
- Output → CLIP Text Encode → Your image generation pipeline

Result: Prompt with professional golden hour lighting!

📄 License

MIT License - Free for commercial and personal use

Need Help?

Check INSTALLATION.md for setup issues
See README.md for full API documentation
Open an issue on GitHub: https://github.com/MushroomFleet/ComfyUI-Variationator

Ready to use?
Start with the Lighting Modifier node and modifier #7 (golden hour)! 🌅

Citation

https://github.com/MushroomFleet/ComfyUI-Variationator

📚 Citation

Academic Citation

If you use this codebase in your research or project, please cite:

@software{comfyui_variationator,
  title = {ComfyUI Variationator: automating image prompt variations},
  author = {[Drift Johnson]},
  year = {2025},
  url = {https://github.com/MushroomFleet/ComfyUI-Variationator},
  version = {1.0.0}
}

15 KiB Raw Permalink Blame History

ComfyUI Variationator - Prompt Orchestration Nodes

📦 Installation

🎯 What This Does

📦 Available Nodes (9 Total)

1. Lighting Modifier (Single)

2. Batch Variation Generator

3. Style Transform

4. Smart Variation (NLP)

5. Preset Collection

6. Modifier Info

7. Prompt Combiner

8. Random Modifier

9. Multi-Style Layer

🌟 The 20 Lighting Modifiers

DRY Category (1-16): Clear Conditions

WET Category (17-20): Dramatic Wet Conditions

🔧 Common Workflow Patterns

Pattern 1: Single Image with Lighting

Pattern 2: Batch Variations

Pattern 3: Style + Lighting Combo

Pattern 4: Smart Creative Workflow

Pattern 5: Preset Collection for Comparison

Pattern 6: Random Exploration

💡 Usage Tips

Best Practices

Modifier Selection Guide

Performance Tips

📋 Example Workflows

Example 1: Golden Hour Car Photography

Example 2: Cyberpunk Style Variation

Example 3: Multi-Variation Comparison

Example 4: Smart Natural Language

Example 5: Random Daily Inspiration

🔄 Integration with Other Nodes

With Wildcard Nodes

With Lora/Checkpoint Switching

With Conditioning Combine

🎨 Advanced Techniques

Technique 1: Progressive Lighting

Technique 2: Style Layering

Technique 3: Conditional Lighting

Technique 4: Batch A/B Testing

📊 Node Reference Quick Chart

🐛 Troubleshooting

Issue: Node outputs empty text

Issue: Batch outputs are all the same

Issue: Style not applying

Issue: Smart Variation not working as expected

🎓 Learning Path

💾 Workflow JSON Examples

🔗 Related Resources

📝 Version Info

🚀 Quick Start Workflow

📄 License

Citation

📚 Citation

Academic Citation

Donate:

15 KiB

Raw Permalink Blame History