How can I composite a live action person inside of a 3d box?

I have been working on a project that has a crate shaped object (you can see through slots) falling over a person. The person is in a background movie which i’ve motioned tracked. Where I’m stuck is how can I get the person inside the box as it falls over them.

So far what I’ve tried is rotoscoping the person, then putting that exported movie onto an image plane which sits inside the box. I’m not sure if this is the right approach b/c i can’t get the proportions to line up over the background in front of the camera yet by eyeballing it.

Any tips or advice would be appreciated!

  • SB

I am not sure about the process in blender as I don’t typically do any masking or compositing in it. I use a different compositing and video software for any vfx or editing. Is the said video of the guy done with a green screen? If not then the method I would go for would be to actually render out the motion of the box with an alpha transparent background so when it is done all you have left is the box movement. Then I would just overlay it onto the video and mask the portions of the box that need to reveal the guy through the slots.

Thanks for your reply!

The person in the video is not shot on a greenscreen, but I did rotoscope them out of the movie with an alpha background (i used after effects for that part <3 = Rotobrush!). I also rendered a clip of just the box falling with an alpha background so i can use either approach.

Method #1
Issue with putting the person on an image plane in the falling box is can’t figure out how to align the image plane to the camera so it overlays the original video. If I could figure that out it’d be ez & done.

Method #2
Issue with overlaying the falling box onto the video is just that, well, i didn’t mention it but the box falls in pieces downwards… like 200+ pieces… it’ll just be [hellishly] tedious to mask.

Thanks for your advice. sounds like your approach is what i need to try.

just following up here - that worked great!
thanks again

1 Like

Glad to hear it worked out for you :slight_smile: And now that I think about it, if the 3d box was a rendered animation then there shoudn’t be any roto work needed for masking as it will just overlay the original video/person therefore being ontop of the person. The open sections of the box falling past the person should cover and uncover the person naturally. The motion track portion of it though as I said I haven’t worked too much in blenders compositor and matchmoving. My typical workflow is to create the camera solve in blender then export it along with the 3d rendered object/alembix/fbx file into Hitfilm Pro which supports 3d modeling and camera solves. There I could just parent the box to one of the imported tracking points/nulls and that portion is done. Then I could composite in shadows and additional vfx.
So now that I look back on this I think just exporting the motion tracked rendered animation with a transparent background with supported alpha you could just alpha over node the 3d box over the video footage of the guy and not have to mess with any roto work.

thanks for the follow up. this has been a great learning project for what blender can do for VFX