Apply ControlNet

💡

This documentation is for the original Apply ControlNet(Advanced) node. The earliest Apply ControlNet node has been renamed to Apply ControlNet(Old). While you may still see the Apply ControlNet(Old) node in many workflow folders you download from comfyui.org for compatibility reasons, you can no longer find the Apply ControlNet(Old) node through search or node list. Please use the Apply ControlNet node instead.

Apply ControlNet

This node applies a ControlNet to a given image and conditioning, adjusting the image’s attributes based on the control network’s parameters and specified strength, such as Depth, OpenPose, Canny, HED, etc.

Documentation

Class name: ControlNetApply
Category: conditioning
Output node: False

Using controlNet requires preprocessing of input images. Since ComfyUI initial nodes do not come with preprocessors and controlNet models, please first install ContrlNet preprocessors download preprocessors here and corresponding controlNet models.

Input Types

Parameter	Data Type	Function
`positive`	`CONDITIONING`	Positive conditioning data, from CLIP Text Encoder or other conditioning inputs
`negative`	`CONDITIONING`	Negative conditioning data, from CLIP Text Encoder or other conditioning inputs
`control_net`	`CONTROL_NET`	The controlNet model to apply, typically input from ControlNet Loader
`image`	`IMAGE`	Image for controlNet application, needs to be processed by preprocessor
`vae`	`VAE`	Vae model input
`strength`	`FLOAT`	Controls the strength of network adjustments, value range 0~~10. Recommended values between 0.5~~1.5 are reasonable. Lower values allow more model freedom, higher values impose stricter constraints. Too high values may result in strange images. You can test and adjust this value to fine-tune the control network’s influence.
`start_percent`	`FLOAT`	Value 0.000~1.000, determines when to start applying controlNet as a percentage, e.g., 0.2 means ControlNet guidance will start influencing image generation at 20% of the diffusion process
`end_percent`	`FLOAT`	Value 0.000~1.000, determines when to stop applying controlNet as a percentage, e.g., 0.8 means ControlNet guidance will stop influencing image generation at 80% of the diffusion process

Output Types

Parameter	Data Type	Function
`positive`	`CONDITIONING`	Positive conditioning data processed by ControlNet, can be output to next ControlNet or K Sampler nodes
`negative`	`CONDITIONING`	Negative conditioning data processed by ControlNet, can be output to next ControlNet or K Sampler nodes

💡

If you want to use T2IAdaptor style models, please use the Apply Style Model node instead

ComfyUI ControlNet Usage Examples

Visit the following pages for examples:

ControlNet Stage Control Settings

In the node settings, you can see two parameters start_percent and end_percent. These parameters can be used to control the application stage of ControlNet during the generation process. When using ControlNet:

You can first set start_percent and end_percent to the default values of 0.000 and 1.000, then adjust these values as needed to see the application effect

Below is a diagram explaining the stage control:

1. Parameter Configuration Reference for Different ControlNet Types

Type	Recommended Weight	Stage Range	Key Preprocessing Parameters	Best Use Cases	Special Techniques
Canny	0.8-1.2	0.0-0.4	Threshold:100/200, Sharpen 15%	Architecture/Product Design	Enable Invert for transparent materials, process complex structures in segments
HED	0.6-0.9	0.2-0.7	Gaussian Blur σ=1.5, Smooth 20%	Portrait/Fashion Design	Anime mode for cartoon style, Realism mode for authentic details
MLSD	0.7-1.0	0.3-0.8	Min Line Length 15px, Angle Tolerance 15°	Engineering Drawing	Increase weight by 0.2 for tilted walls, decrease by 0.3 for glass curtain walls
Depth	0.7-1.0	0.2-0.9	MiDaS Large Model, 3D Mapping	VR/Medical Visualization	Near-view enhancement mode for subject details, ZoeDepth for macro scenes
Normal	0.5-0.8	0.4-1.0	Resolution 2048px, AO 0.3	Product Rendering	Enable Specular for metal materials, multi-light synthesis for enhanced 3D feel
Scribble	0.4-0.7	0.5-1.0	SoftEdge Blur 3px, Hue Tolerance 15%	Concept Design	Use 50% opacity masks for gradients, Pantone library for brand consistency
Lineart	0.6-0.9	0.3-1.0	Anti-aliasing On, Line Width ±2px	Character Art	Anime mode for simplified lines, Realism mode for complex folds
OpenPose	0.9-1.1	0.0-0.3	25-point Skeleton, Hand Detail Enhancement	Motion Capture	Motion blur compensation to prevent ghosting, increase weight to 1.2 for martial arts
Segmentation	0.8-1.0	0.0-0.7	ADEPT 2.0, Mask Feather 10px	Ad Composition	Decrease weight to 0.2 for sky regions, sharpen building edges by 20%
Tile	0.3-0.6	0.4-0.9	256x256 Blocks, 30% Repeat Rate	Texture Generation	Randomize variations for natural feel, enable seamless tiling for brick walls

2. Classic Scene Configuration Templates

2.1 Architectural Visualization Design

Control Type	Weight	Stage Range	Preprocessing Parameters	Adjustment Tips
Canny	1.0	0.0-0.4	Threshold 100/200	Enable Invert for glass walls
Depth	0.8	0.2-0.7	MiDaS Large Model	Enhance mid-ground by 20%
MLSD	0.6	0.5-0.9	Min Line Length 20px	Increase weight to 0.8 for tilted walls

2.2 Game Character Design

Control Type	Weight	Stage Range	Preprocessing Parameters	Dynamic Adjustment
OpenPose	1.0	0.0-0.3	Full Skeleton	Reduce to 0.7 after step 20
Lineart	0.7	0.4-1.0	Anime Mode	+0.1 weight for equipment areas
Scribble	0.5	0.5-1.0	SoftEdge Blur 2px	Set color block boundary strength to 0.3

2.3 Product Concept Design

Control Type	Weight	Stage Range	Preprocessing Parameters	Material Optimization
HED	0.9	0.0-0.3	Gaussian Blur σ=1.5	Enable Specular for metal surfaces
Normal	0.7	0.2-0.6	Resolution 2048x2048	Reduce to 0.5 for plastic materials
Depth	0.6	0.5-0.9	Near View Enhancement	Background blur strength 1.2

2.4 Medical Visualization

Control Type	Weight	Stage Range	Preprocessing Parameters	Precision Control
Scribble	0.8	0.0-0.5	Red Annotation Lines	Organ boundary tolerance ±2px
Depth	0.7	0.4-0.8	CT Scan Mode	Layer spacing 0.1mm
Lineart	0.9	0.7-1.0	Ultra Detail	Blood vessel path precision 1px

2.5 Film Scene Composition

Control Type	Weight	Stage Range	Preprocessing Parameters	Atmosphere Creation
Seg	0.9	0.0-0.6	ADEPT Model	Reduce sky region weight to 0.2
Shuffle	0.6	0.3-0.8	Color Temp 5500K	Neon light area weight 0.8
Depth	0.7	0.5-1.0	Dynamic Range Compression	Foreground sharpening 1.5

2.6 E-commerce Advertisement Design

Control Type	Weight	Stage Range	Preprocessing Parameters	Commercial Optimization
Canny	1.2	0.0-0.4	Edge Sharpen +15%	Enhanced reflection mode
Scribble	0.7	0.3-0.7	Pantone Library	Brand color tolerance ±5%
Inpaint	0.5	0.6-1.0	Feather Radius 15px	Text area protection mask

3. Expert-Level Adjustment Strategies

3.1 Stage Weight Decay Model

Generation Progress	Control Type	Decay Curve	Formula Example
0-30%	Structure Control	Constant Strength	strength = 1.0
30-70%	Spatial Control	Linear Decay	strength = 1.0 - (step-30)/40*0.5
70-100%	Detail Control	Reverse Enhancement	strength = 0.5 + (step-70)/30*0.5

3.2 Multi-ControlNet Conflict Resolution

Conflict Type	Visual Manifestation	Resolution Strategy
Structure-Space	Object floating/perspective errors	Set stage interval ≥0.15
Space-Detail	Material distortion/reflection anomalies	Add area masks to isolate control ranges
Structure-Detail	Key feature loss	Increase structure control strength by 20%

4. Common Issues Quick Reference

Q1: Control effect suddenly disappears?
✅ Check if end_percent ends too early (recommended ≥0.8)
✅ Confirm no other ControlNet overlaps in the area

Q2: Generation results show ghosting?
✅ Reduce stage overlap (recommended ≤20%)
✅ Set exclusion masks for conflicting ControlNets

Q3: How to optimize for insufficient VRAM?
✅ Use stepped stage configuration (example: 0.0-0.3 → 0.4-0.6 → 0.7-1.0)
✅ Reduce non-critical ControlNet resolution to 512px

Model Resources: controlNet Model Resources Download
Preprocessor Plugin: ComfyUI ControlNet Auxiliary Preprocessors