ComfyUI for Nuke generates Depth Map
I tested using ComfyUI for Nuke with DepthAnything to generate a depth map from footage. The biggest issue I encountered was that the footage was in .dpx sequences. ComfyUI’s built-in Load Image node cannot read .dpx or .exr files, so I downloaded ComfyUI-HQ-Image-Save, which includes a LoadEXR node that can read .dpx files.During parameter passing, errors kept occurring. The problem was that LoadEXR requires a tone map parameter, but the Nuke Read node does not provide one. My solution was to modify the source code and hardcode the tone map parameter to “linear,” which allowed it to run successfully. 【 Reflection 】 The accuracy of DepthAnythingV2 is slightly insufficient, but the model is small and generates results quickly. In the future, I could try other depth map generation models such as MiDaS. For tasks like adding depth of field or some fog effects, it seems acceptable, but it is not quite suitable for relighting. 【 References 】 [https://youtube.com/watch?v=EA_xvTTjpYA]