The Depthflow node is central to the node pack, designed to take an image (or video) along with its corresponding depth map. It applies various motion animations—such as Zoom, Dolly, and Circle—to create a parallax effect, ultimately producing a batch of images that can be rendered as a video.

In this context, the depth model outputs metric depth, indicating the distance from the camera in meters, which is essential for 3D projection. To convert this metric depth into a relative scale suitable for ControlNet, it needs to be remapped to a 0 to 1 range using a separate node. Default settings typically work well, but adjustments can be made for brightness through inversion or gamma. If you encounter errors related to "vit_large_patch14_dinov2," make sure to update the timm library to versions 0.9.16 or 1.0.9.

ComfyUI-Depth-Pro

https://github.com/spacepxl/ComfyUI-Depth-Pro

ComfyUI-Depthflow-Nodes

https://github.com/akatz-ai/ComfyUI-Depthflow-Nodes

model Dowmload:

ComfyUI/models/depth/ml-depth-pro/

https://huggingface.co/spacepxl/ml-depth-pro/blob/main/depth_pro.fp16.safetensors

If you’re loving my workflow or it help you a little bit, don’t shy to sprinkle some love my way with a ❤️ ! Your support means the world!

____________________

ComfyUI

https://www.facebook.com/groups/comfyui