ComfyUI Node Runs on cloud
Run Depth Anything 3
Run Depth Anything 3 on an image. In multi-view mode each image is treated as a separate view of the same scene.
Run Depth Anything 3
- da3_model
- image
- da3_geometry
◄resolution504►
◄resize_methodupper_bound_resize►
◄mode▾►
Categoryimage/geometry estimation
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| da3_model | DA3_MODEL | — | |
| image | IMAGE | — | |
| resolution | INT | 504140–2520 | Resolution the model runs at (longest side, multiple of 14). Lower = faster / less VRAM. Higher = more detail. Output is upsampled back to the original size. |
| resize_method | COMBO | upper_bound_resize | upper_bound_resize: scale so the longest side = resolution (caps memory, default). lower_bound_resize: scale so the shortest side = resolution (preserves more detail on tall/wide images, uses more memory). |
| mode | COMBO | mono: single view image (works with any model variant). multiview: all images processed together for geometric consistency + camera pose (for Small/Base models only). |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| da3_geometry | DA3_GEOMETRY | Dictionary of non-normalized tensors. Always has the keys: depth, image, mode. Optional keys: sky (for Mono/Metric), confidence (for Small/Base), extrinsics + intrinsics (for multi-view). |