ControlNet - OpenPose

Learn to use pose detection ControlNet.

What is OpenPose

OpenPose detects human body keypoints, allowing pose control in generation.

Installation

Download Model

ComfyUI/models/controlnet/
  control_v11p_sd15_openpose.pth

Preprocessor Models

comfyui_controlnet_aux/
  models/openpose/

Keypoint Detection Types

Body Keypoints

Keypoint	Description
Nose	Head position
Eyes, Ears	Face direction
Shoulders	Upper body
Elbows, Wrists	Arms
Hips	Lower body
Knees, Ankles	Legs

Additional Detection

Option	Description
Face	Facial features
Hands	Hand keypoints
Whole Body	Full detection

Workflow

[LoadImage] → [OpenPosePreprocessor] → [ControlNet] → [Apply ControlNet]
                                    ↑
[CLIPTextEncode] → [KSampler] ← [CLIPTextEncode]

Parameters

OpenPose Preprocessor

Parameter	Description
Detect	body / face / hand / full
Resolution	Detection quality

ControlNet Settings

Parameter	Recommended
Strength	0.6-0.85
Start	0.0
End	0.7

Use Cases

Character Poses

# Prompt
beautiful woman, standing pose, elegant dress, detailed

# Input
OpenPose from reference image

Action Scenes

# Prompt
person dancing, dynamic pose, flowing dress, magical

# Input
Dance pose reference

Fashion

# Prompt
model wearing summer dress, fashion photography

# Input
Fashion pose reference

Creating Pose References

Method 1: Photo

1. Take photo of person in desired pose
2. Run through OpenPose preprocessor
3. Use detected keypoints

Method 2: Rendered Image

Use 3D pose software:

• Blender
• DAZ Studio
• Poser

Method 3: Manual

Draw skeleton with keypoints (advanced).

Tips

1. Clear pose: Use simple, clear references
2. Multiple poses: Combine different detections
3. Strength adjustment: Lower if too rigid

Common Issues

Issue	Solution
Missing keypoints	Use higher resolution
Wrong pose	Check source image quality
Too stiff	Lower end parameter

Next Steps

• Multi-ControlNet - Combine with other ControlNets
• Canny Control - Edge guidance
• Depth Control - Structure control