WORKFLOW · #depth · low vram
Depth Anything V2
A clean monocular depth map from any image — for 2.5D parallax, depth-of-field, fog, and depth-conditioned ControlNet.
Before / After
imagedepthWorkflow Preview
How It Works
Depth Anything V2 estimates a dense depth map from a single image — no stereo, no LiDAR. The result drives 2.5D camera moves, synthetic depth-of-field and atmospheric fog, or conditions a depth ControlNet for structure-locked generation. Fast and robust across faces, scenes and product shots.