Regional Conditioning for ComfyUI

The simplest way to control exactly where things appear in your images.

Draw boxes on a canvas, type what you want in each region, and generate! Perfect for precise composition control in Stable Diffusion, SDXL, Flux, and Chroma.

---

📸 Example Workflow

Example workflow showing the simple 3-node setup: Checkpoint → CLIP → Regional Prompter → Sampler

---

🚀 Quick Start (60 seconds)

For Flux / Chroma (Modern Models)

1. Load your Flux/Chroma checkpoint
2. Add the 🎨 Regional Prompter (Flux/Chroma - Easy!) node
3. Connect CLIP from your checkpoint to the node
4. Type in the background prompt: "photo of a city street"
5. Draw a box on the canvas where you want something specific
6. Type in Region 1 prompt: "red sports car"
7. Draw another box for something else
8. Type in Region 2 prompt: "street vendor"
9. Connect the conditioning output to your sampler
10. Generate! 🎉

That's it! No external nodes, no complex wiring. Just prompts and boxes.

For SDXL / Stable Diffusion

Same steps, but use 🎨 Regional Prompter (SD/SDXL - Easy!) instead.

---

📥 Sample Workflows (Download & Import)

Ready-to-use workflow JSON files - just drag and drop into ComfyUI!

• Flux Workflow - Flux.1-dev with 2 regions, optimized settings
• SDXL Workflow - SDXL with 3 regions, area-based conditioning
• Qwen-Image Workflow - Experimental Qwen-Image setup (untested)

Don't have these files yet? Check the examples folder or create your own using the Quick Start guide above!

---

✨ Why Use This?

Without Regional Conditioning:

• "A tiger and a berry bush in front of mountains"
• → AI puts them wherever it wants (usually random placement)

With Regional Conditioning:

• Background: "Mountains"
• Box 1 (left side): "Tiger"
• Box 2 (right side): "Berry bush"
• → AI puts them exactly where you drew the boxes

Perfect for:

• Character positioning in scenes
• Multi-subject compositions
• Product photography layouts
• Architectural visualization
• Comic/storyboard panels

---

📦 Installation

Method 1: ComfyUI Manager (Easiest)

1. Open ComfyUI Manager
2. Search for "Regional Conditioning"
3. Click Install
4. Restart ComfyUI

Method 2: Manual

cd ComfyUI/custom_nodes/
git clone https://github.com/EnragedAntelope/ComfyUI_RegionalConditioning
# Restart ComfyUI

---

🎨 Which Node Should I Use?

🌟 RECOMMENDED: Enhanced Easy Nodes

These are the easiest way to use regional conditioning:

🎨 Regional Prompter (Flux/Chroma - Easy!)

• ✅ For: Flux, Chroma, SD3, SD3.5
• ✅ Type prompts directly in the node
• ✅ Optimized for Flux (perfect mask strength and blending)
• ✅ Just connect CLIP and go!

🎨 Regional Prompter (SD/SDXL - Easy!)

• ✅ For: Stable Diffusion 1.5, SD 2.x, SDXL
• ✅ Type prompts directly in the node
• ✅ Same simple workflow
• ✅ Perfect for traditional SD models

Use these unless you have a specific reason not to.

🔧 Advanced Nodes (For Power Users)

If you need more control or want to share CLIP encodings across multiple nodes:

• Multi Area Conditioning (SD/SDXL - Advanced) - Area-based for traditional SD
• Multi Area Conditioning (Flux/Chroma - Advanced) - Mask-based for modern models

These require external CLIP Text Encode nodes but offer more flexibility.

---

🎯 Model Compatibility

Model Type	Node to Use	Status	Notes
Stable Diffusion 1.5	🎨 Regional Prompter (SD/SDXL)	✅ Fully Supported	No region limit
Stable Diffusion 2.x	🎨 Regional Prompter (SD/SDXL)	✅ Fully Supported	No region limit
SDXL	🎨 Regional Prompter (SD/SDXL)	✅ Fully Supported	No region limit
Flux (all variants)	🎨 Regional Prompter (Flux/Chroma, etc.)	✅ Fully Supported & Optimized	3-4 regions max recommended
Chroma (Radiance, etc.)	🎨 Regional Prompter (Flux/Chroma, etc.)	✅ Architecture Supported	Likely 3-4 region limit (untested)
SD3 / SD3.5	🎨 Regional Prompter (Flux/Chroma, etc.)	✅ Architecture Supported	No known region limit
Qwen-Image	🎨 Regional Prompter (Flux/Chroma, etc.)	🟡 Experimental	Try it and report results!

---

💡 How to Use (Detailed)

Understanding the Canvas

When you add a Regional Prompter node, you'll see:

• A canvas (black rectangle) - This is your drawing area
• Prompt boxes (text inputs) - Background + up to 4 regions
• Drawing controls (below canvas) - Position and size your boxes

Drawing Your First Region

1. Start with background prompt:

   • Type: "photo of a sunny beach"
   • This applies to the whole image

2. Add your first region:

   • In Region 1 prompt, type: "woman in red dress"
   • Below the canvas, find the controls for Region 1
   • Set: x=100, y=200, width=300, height=400
   • You'll see a colored box appear on the canvas!

3. Adjust the box visually:

   • Change the index dropdown to select which region to edit
   • Each region gets a unique color
   • The 64px grid helps you align precisely

4. Add more regions:

   • Type in Region 2 prompt: "beach umbrella"
   • Adjust x/y/width/height to position it
   • Repeat for up to 4 regions total

5. Generate!

   • Connect to your sampler and generate
   • Each prompt appears exactly where you drew it

Tips for Best Results

For Flux Users:

• ✅ Keep "Soften Masks" ON (it's on by default) - Uses 0.8 strength that works better than full strength
• ✅ Use 3-4 regions maximum - More regions = quality degradation
• ✅ Increase CFG to 5-7 - Flux needs higher guidance for regional control
• ✅ Draw larger boxes - Bigger regions = better control

For Flux-Based Models (Chroma, etc.):

• ✅ Keep "Soften Masks" ON - Works well with Flux-based architectures
• ⚠️ Likely 3-4 region limit - Chroma is based on Flux, may have same limitations (untested)
• ✅ Start with Flux settings - Increase CFG to 5-7, try and adjust

For SDXL Users:

• ✅ Align to 64px grid - Matches latent space boundaries
• ✅ No region limit - Use as many as you need
• ✅ Normal CFG values - 7-9 works great

General Tips:

• Overlapping regions are processed in order (Region 1 → 2 → 3 → 4)
• Strength slider (0.0-10.0) controls region influence:
  • 1.0 = Normal (default)
  • 0.5 = Subtle hint
  • 2.0+ = Strong emphasis
  • 0.0 = Effectively disabled
• Leave prompts empty to disable unused regions
• Fullscreen regions (x=0, y=0, width=canvas, height=canvas) act as background

---

📚 All Available Nodes

Enhanced (Easy Mode) - RECOMMENDED

🎨 Regional Prompter (SD/SDXL - Easy!)

Perfect for Stable Diffusion and SDXL

Inputs:

• clip - CLIP model from your checkpoint
• background_prompt - Overall scene description (multiline text)
• region1_prompt through region4_prompt - What appears in each box (optional)

Outputs:

• conditioning - Connect to your sampler
• width, height - Canvas dimensions (for reference)

How it works: Type prompts directly, draw boxes, done! The node handles all CLIP encoding internally.

---

🎨 Regional Prompter (Flux/Chroma - Easy!)

Perfect for Flux, Chroma, and SD3

Inputs:

• clip - CLIP model from your checkpoint (from Load Checkpoint node)
• width, height - Output image dimensions (just numbers - no latent needed!)
• background_prompt - Overall scene description (multiline text)
• region1_prompt through region4_prompt - What appears in each box (optional)
• soften_masks - Soften region edges (default: ON - RECOMMENDED!)

Outputs:

• conditioning - Connect to your KSampler's positive conditioning input

How the Workflow Connects:

Load Checkpoint → CLIP
                   ↓
    Regional Prompter (Flux/Chroma)
    (Set width/height here, e.g., 1024x1024)
                   ↓
              conditioning
                   ↓
            KSampler (positive input)

💡 About "Soften Masks" (IMPORTANT!):

• ✅ KEEP ON (default): Softer region edges, better blending (0.8 strength + gentle feather)
• ❌ Turn OFF only if: You want sharp, harsh region boundaries (1.0 strength, no feather)
• Why it's better: Research shows softer masks (0.8) produce better results than full-strength (1.0) for Flux
• For other models: Works well with mask-based models - try it ON first, disable only if you prefer sharper edges
• What feathering does: Gentle edge blending (5-8 pixels) to avoid harsh boundaries - you can disable by turning this toggle OFF

💡 Pro Tip - Auto Width/Height: You can connect the width/height outputs from your Empty Latent Image node directly to this node's width/height inputs! No need to type them manually - they'll stay in sync automatically.

---

Advanced Nodes (For Power Users)

Multi Area Conditioning (SD/SDXL - Advanced)

Area-based conditioning for traditional Stable Diffusion models. Requires external CLIP Text Encode nodes.

Inputs:

• conditioning0 - Base/background conditioning (connect CLIPTextEncode here)
• conditioning1, conditioning2, etc. - Regional conditioning (add more via right-click menu)

Outputs:

• conditioning - Combined regional conditioning
• resolutionX, resolutionY - Canvas dimensions

Right-Click Menu:

• Insert input above/below current
• Swap with input above/below
• Remove currently selected input
• Remove all unconnected inputs

---

Multi Area Conditioning (Flux/Chroma - Advanced)

Mask-based conditioning for modern models. Requires external CLIP Text Encode nodes.

Inputs:

• conditioning0 - Base/background conditioning
• conditioning1, conditioning2, etc. - Regional conditioning
• width, height - Output dimensions
• mask_strength - Mask intensity (0.7-0.85 recommended for Flux, 1.0 for others)

Output:

• conditioning - Combined mask-based conditioning

How it works: Boxes are automatically converted to binary masks. You draw visually, the node handles mask generation.

---

Utility Nodes

Conditioning Upscale

Scale conditioning areas by a multiplier - perfect for hi-res fix workflows.

Example: 512x512 regions × scalar=2 → 1024x1024 regions

---

Conditioning Stretch

Resize conditioning to fit new dimensions - more flexible than upscale.

Example: Transform regions from 512x512 to 1024x768 (proportional scaling)

---

Multi Latent Composite (Visual)

Visual interface for compositing multiple latents with positioning and feathering.

Features:

• ✅ Automatic bounds checking and clipping
• ✅ Channel compatibility validation
• ✅ Smart feathering with gradient masks
• ✅ Detailed warnings and info messages

---

🔧 Advanced Techniques

Fullscreen/Background Conditioning

Set a region to cover the entire canvas (x=0, y=0, width=canvas_width, height=canvas_height) to make it fullscreen. This is perfect for background prompts that should influence the whole image.

Grid Alignment

The 64px grid overlay helps you align regions to latent space boundaries (VAE works in 8x downscaled space, 64px = 8 latent pixels).

Strength Control

Every region has a strength slider (0.0-10.0):

• 1.0 - Normal strength (default)
• 0.5 - Subtle influence (region blends more with background)
• 2.0+ - Strong influence (region dominates)
• 0.0 - Effectively disabled

Overlapping Regions

Regions are processed in order. If boxes overlap:

• Background applies first (fullscreen)
• Region 1 applies on top
• Region 2 applies on top of Region 1
• And so on...

The last region "wins" in overlapping areas (for mask-based conditioning).

Dynamic Sizing

Important: The canvas size can be different from your output size!

For example:

• Draw on a 512x512 canvas
• Generate at 1024x1024 output
• Regions automatically scale 2x

This makes it easy to design compositions at lower resolution, then generate at higher resolution.

---

❓ Troubleshooting

"Workflow metadata missing" error

Cause: ComfyUI hasn't saved your workflow yet Fix: Save the workflow (Ctrl+S or Cmd+S), then reload the page

Canvas doesn't show boxes I drew

Cause: Workflow not saved, or coordinates are outside canvas Fix:

1. Save and reload workflow
2. Check that x+width ≤ canvas width and y+height ≤ canvas height
3. Select the region using the index dropdown to see where it is

Flux regions look harsh or over-defined

Cause: Flux Optimize is disabled, or using advanced node with mask_strength=1.0 Fix:

• Use the Easy node with Flux Optimize ON
• Or set mask_strength to 0.7-0.85 in the advanced node

Too many regions = bad quality (Flux)

Cause: Flux struggles with more than 4 regions Fix: Combine related elements into single regions, or use fewer boxes

Region has no effect

Cause: Prompt is empty, strength is 0.0, or box is too small Fix:

1. Make sure the region prompt has text
2. Check strength slider is at least 0.5
3. Make boxes at least 64x64 pixels

---

📝 Changelog

Version 3.1 (November 22, 2025)

• ✨ NEW: Enhanced "Easy" nodes with inline prompts!
  • 🎨 Regional Prompter (SD/SDXL - Easy!)
  • 🎨 Regional Prompter (Flux/Chroma - Easy!)
• ✨ NEW: No more external CLIP Text Encode nodes needed
• ✨ NEW: Flux-optimized mask strength (0.8) and feathering
• ✨ NEW: Up to 4 regions + background in one node
• 📚 DOCS: Completely rewritten README for user-friendliness
• 📚 DOCS: Added CONTEXT.md for development learnings

Version 3.0 (November 22, 2025)

• ✨ NEW: MultiAreaConditioningMask for Flux/Chroma/SD3 support
• ✨ NEW: Comprehensive error handling and validation
• ✨ NEW: Bounds checking for MultiLatentComposite
• ✨ NEW: Tooltips and descriptions on all inputs
• 🔧 FIX: Modernized JavaScript loading (no more file copying!)
• 🔧 FIX: Removed debug console.log statements
• 🔧 FIX: Fixed graph reference bug in background drawing
• 📚 DOCS: Complete compatibility matrix and examples
• 🚀 PERF: Improved tensor operations and validation

Version 2.4 (Original)

• Visual area conditioning interface
• MultiLatentComposite with feathering
• ConditioningUpscale and ConditioningStretch utilities

---

🤝 Contributing

Found a bug? Have a feature request? Open an issue on GitHub!

Tested with:

• ComfyUI 0.3.71+
• Stable Diffusion 1.5, SD 2.x, SDXL
• Flux.1-dev, Flux.1-schnell
• Chroma1-Radiance

---

📄 License

GLWT (Good Luck With That) Public License

See LICENSE file for the full text.

TL;DR: Use it however you want, but don't blame us if it breaks! 😄

---

🙏 Credits

Current Maintainer: EnragedAntelope Repository: github.com/EnragedAntelope/ComfyUI_RegionalConditioning

Original Author: Davemane42 Original Concept: Visual regional conditioning interface for ComfyUI

Modernization & Enhancement (2025):

• Flux/Chroma/SD3 support with mask-based conditioning
• Enhanced "Easy Mode" nodes with inline prompts
• Comprehensive error handling and validation
• User-friendly documentation rewrite

Special Thanks:

• ComfyUI community for testing and feedback
• Black Forest Labs for Flux research and documentation
• All contributors who reported issues and suggested improvements

---

📸 Complete Workflow Examples

Easy Mode Workflow (RECOMMENDED)

The simplest possible setup - only 3 nodes needed!

Load Checkpoint
    ├─→ CLIP ─→ Regional Prompter (Flux/Chroma - Easy!)
    │             ├─ Set width: 1024
    │             ├─ Set height: 1024
    │             ├─ background_prompt: "your scene description"
    │             ├─ region1_prompt: "what goes in box 1"
    │             ├─ region2_prompt: "what goes in box 2"
    │             └─ soften_masks: ✅ ON (recommended!)
    │
    ├─→ MODEL ──┐
    └─→ VAE ───┐│
               ││
    Regional Prompter → conditioning
               ││
               ↓↓
            KSampler
         (positive: conditioning from Regional Prompter)
         (negative: empty or your negative prompt)
         (model: from Load Checkpoint)
         (latent: Empty Latent Image at matching size)
               ↓
          VAE Decode
               ↓
          Save Image

Key Points:

• ✅ Width/height in the Regional Prompter should match your Empty Latent Image size
• 💡 PRO TIP: Connect width/height outputs from Empty Latent Image → Regional Prompter for auto-sync!
• ✅ No latent input needed on the Regional Prompter - it just needs the numbers!
• ✅ The node outputs CONDITIONING, not an image - connect to KSampler's positive input
• ✅ Keep "Soften Masks" ON for better results (especially Flux and Flux-based models)

Advanced Mode Workflow (For Power Users)

When you need shared CLIP encodings or more control:

Load Checkpoint → CLIP → CLIPTextEncode (background)
                            ↓
                     CLIPTextEncode (region 1)
                            ↓
                     CLIPTextEncode (region 2)
                            ↓
            Multi Area Conditioning (collects all)
                            ↓
                   conditioning output
                            ↓
                        KSampler → VAE Decode → Save Image

When to use Advanced Mode:

• You want to share CLIP encodings across multiple nodes
• You need manual control over mask strength per region
• You're doing complex prompt engineering with multiple text encode nodes

---

Enjoy precise regional control in your ComfyUI workflows! 🎨

Questions? Check the troubleshooting section above, or open an issue on GitHub.