ControlNet - Depth Map

Learn to use depth map ControlNet.

What is Depth ControlNet

Uses depth map to provide 3D structure information, maintaining spatial relationships.

Installation

Download Model

ComfyUI/models/controlnet/
  control_v11f1p_sd15_depth.pth

Preprocessor

Uses MiDaS for depth estimation:

comfyui_controlnet_aux/
  models/midas/

Workflow

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

Parameters

Depth Preprocessor

Preprocessor	Description
MiDaS (DPT-Large)	High quality, slower
MiDaS (DPT-Small)	Faster, less detailed
LeReS	Good for outdoors

ControlNet Settings

Parameter	Recommended
Strength	0.6-0.9
Start	0.0
End	1.0

Use Cases

3D Scene Generation

# Prompt
cozy living room, modern furniture, warm lighting, interior design

# Input
Reference image or manually created depth

Character Composition

# Prompt
beautiful woman standing in garden, detailed, 8k

# Input
Depth map from pose reference

Architecture

# Prompt
modern architecture, concrete and glass, minimalist design

# Input
Depth from reference or sketch

Creating Depth Maps

Method 1: From Photo

Use preprocessor to generate from existing image.

Method 2: Manual Creation

Create depth map manually (darker = further).

Method 3: Using LoRA

Train depth LoRA for consistent depth patterns.

Tips

1. Depth quality: Use high quality source images
2. Combine with canny: Can use both for better structure
3. Adjust strength: Higher for stronger structure influence

Common Issues

Issue	Solution
Flat depth	Check source image quality
Wrong perspective	Verify depth direction
Too controlling	Lower strength

Next Steps

• ControlNet OpenPose - Pose control
• Canny ControlNet - Edge control
• Multi-ControlNet - Combine controls