So there is this great function in nuke that i use all the time which is that i can projekt an image/sequence onto a 3d object, have the scaneline render output the projektet texture as a flattend image using the 3D objects UV’s giving a stabilized image. I use this all the time for cleanup and really missing it in flame. Closest I’ve found is tracking the UV of a plane in Flame, it works good for alot of cases, but in this case i have a 3d tracked head and would like go get the face texture output as a flat image.
Ah, yes…the famous Jet Li Nuke demo. I wish we had this, too!
I believe @DannyYoon has something for this that can work.
But yeah, ever since that Jet Li demo I’ve wanted that tool so bad.
Hmm, i might be way off but my thinking is, wouldnt it be possible, since action can output a object with UV, to have som form of tool that compares the uv and the diffsue(clean projektet output) and map it to a flat image? But this is abit above my level of expertice.
Here’s the UV Unwrap setup which is based on a something Lewis Saunders posted on the logik Facebook group. I haven’t actually used it for awhile…might need an update for UHD resolutions. I’ll take a look in the morning…
3_UV_UnWrap_DY_v01_v01.zip (673.6 KB)
The only issue with unwrapping 3D models with pixels spread is that the UVs can adhere to spacially disparate areas of the geo. Something that’s totally fine if you have the geo and know where the breaks occur on a vertext level but once you have a rendered piece of material that knowledge is gone. So when you have a model like a head you need to make sure that it’s uv’d in such a way that there are no overlaps spacially on the model because as you pixels spread the unwrap flame is basically smearing those pixels from one location to another. So if, hypothetically you see a seam in your UVs or the eyes are uv’d separately, whatever, you’ll end up with overlapping pixels spreading in different directions.
If you have a single, watertight head geo, eyes part of the mesh, UV’d such that you only ever see a continuous increase in value across the visible areas of the mesh then the pixel spread technique will work.
Holes in the mesh will fuck shit up just as much as bad UVs
Wow! this looks really cool! Thanks for sharing Danny