Auto Crop Black Bars

[Schniefel]

I have a video which has different aspect ration, e.g. 16:9 and 21:9 in some scenes.
I don't want to crop the whole video to 21:9 and now I wonder if the black bars in the 21:9 scenes affect 3D generation.

In the 3D generator I see the option to "Auto Crop Black Bars", but i am unsure if this is done per frame or simply for the whole video.
Additionally I would have expected the option also on the depth map generation because I think the black bars would affect the depth calculation also, so simply skipping the black bars in 3D generation would not be sufficient.

Do you have any clearification/recommendations on that topic?

My current idea is to split the video into parts with different aspect ratio, crop the 21:9 parts and separately calculate the depth map and the 3D video of each part and finally join the different 3D parts together.

[VisionDepth]

Sorry for the late reply.

When generating a depth map, VisionDepth3D’s Depth Engine already handles top and bottom black bars by treating those areas as neutral depth so they should not strongly affect the depth shift.

For the 3D Generator, the Auto Crop Black Bars option checks the first frame for top and bottom black bars, caches that crop value, and then applies the same crop throughout the full render. It is not currently detecting and changing the crop per frame.

So if your video changes aspect ratio during the movie, for example some scenes are 16:9 and others are 21:9, the safest workflow is to split the video at the aspect-ratio changes, render each section separately with the correct settings, then join the finished 3D parts together afterward.

One important note: if you crop a 1920x1080 source down to something like 1920x800, make sure your output/aspect settings match that cropped shape. Otherwise the cropped image could be resized back into a 16:9 frame and look vertically stretched.

So your idea is basically the best approach right now:

1. Split the video by aspect-ratio sections
2. Crop/render the 21:9 sections separately
3. Generate matching depth maps for each section
4. Render each section in 3D
5. Join the finished 3D clips together

I may improve this in the future with smarter per-scene or per-frame aspect handling, but for now separate section renders are the safest method.

[Schniefel]

Thank you for your reply. I will try that out. 👍