Depth Map Techniques
A grayscale image representing distance from the camera, where brightness indicates depth. Used with ControlNet to maintain spatial structure during generation.
Depth maps provide 3D spatial information to guide AI image generation.
How Depth Maps Work
- White areas: Close to camera
- Black areas: Far from camera
- Gray gradients: Intermediate distances
Creating Depth Maps
From Photos (Estimation)
- MiDaS, Depth Anything models
- Built into most ControlNet UIs
- Approximate but effective
From 3D Software
- Render depth pass
- Exact depth values
- Perfect for 3D-assisted workflows
Hand-Drawn
- Simple geometric depth
- Rough spatial guidance
- Minimal effort
Using with ControlNet
- Load image or create depth map
- Apply depth preprocessor if needed
- Use Depth ControlNet model
- Set control strength (0.5-1.0)
- Generate with desired prompt
Use Cases
- Scene reconstruction: Maintain room layout
- Character posing: Preserve depth relationships
- Landscape control: Keep foreground/background
- Object placement: Correct spatial ordering
Depth vs Other Controls
| Control | What It Captures |
|---|---|
| Depth | 3D spatial relationships |
| Canny | Edge outlines |
| Pose | Human skeleton |
| Segmentation | Object regions |
Tips
- Combine depth with other controls
- Lower strength for more creativity
- Clean up estimated depth maps
- Works best with clear depth separation