I have been trying to think of ways to paint on polygon models without the burden of uv's. Projection paint in apps such as Deep Paint 3D and Body paint allow for projection painting, but the polygon is limited by it's uv's. Dependence upon uv's while painting models must stop. So far I have tested the following method successfully and have recieved great looking textured models. An automated system should be built to streamline this method for use with the 3d texture paint tool as well.
UV INDEPENDENT POLYGON TEXTURE PAINTING
step 1:
extract and seperate faces. delete history. assign shader with object's name to each new object.
select first object.
step 2:
create temp uvset create uvset with corresponding object name. Layout. copy map1 to temp uvset. copy (object name)'s uvset to map1. delete history.
step 3:
select next object. repeat step 2.
step 4:
PAINT.
step 5.
select object.
step 6.
copy map1 to (object name) uvset. link shader's file node to (object name)'s uvset. copy temp uvset to map1. delete temp uvset. delete history.
step 7.
select next object. repeat step 6.
step 8.
select all objects and combine.
Make the world a better place and help write this script! thanks. Skeletonkey
Joojaa. you're right that it is bulky, but it works very well and surprisingly fast so far. My method of painting models is much easier and allows for much more control. It's like painting on a model where each face is a nurbs patch. "Anyway that system is used by brush..." What does that mean? ZBrush?
what's a catch of painting on bunch of individual faces? how method of Yours could be more productive than existing capabilites of Maya's UV tools and paint? to be honest I didn't understood reasonability of all this face-extract-combine stuff and so on. At present moment with better success You could i) perform automatic mapping, then ii) select all faces and perform individual face uv normalization with preserved aspect ratio (why?) and then iii) layoutUVs - voila! but where's a catch again? Unfortunately for Your method, polygonal faces could be triangular (so they never will be like NURBS) and wery stretched (it means that pixel aspect for each face would be individual). Also.. do You expect to have single texture and/or uvSet for each face? Sounds like a nonsense for model with 10K faces. Combining after all this mess is not enough too - do not forget about they should be MERGED and SOFT/HARD-ened again. Pehaps I am missing a point of Yours, so please explain more - where is advantage of it. Will You?
Well its easier to not to separate faces at all... But to doa palnar on a per face like said.
ANYWAY: theres other things that uv layot does besides cause s stretch and he4adache.
A) it lowers rendertimes, because the filtering can be accoplished in the lookup kernel. Alo its possible to reduce mem usage too with pyramidical mapping. (if you hava lot of semas then this filtering leads to more sampling and taht is allways a bit over the top expensive)
the uvmap can act as a guise for procedural textures.
C) the ability to use the cordinate space better for where thing count most. Also translates to the ability to use 2d prgs tp do this without need for 3d paint wich can be akward at times...
Yes. A shader, texture and uvset for each face along with uvset map1 containing all the faces.
Even if some polygonal faces are triangles, it still works.
Yes I know that the verts must be merged and faces softened or hardened.
I have found that by planar projecting a uvmap (using automatic placement for each face), then laying out the uv's for each seperated object works best....so far.
A 10k face model will produce 10k maps. This is heavy, but it is not neccessary to use this method of mapping throughout the model.
So,....THE POINT: painting a model mapped with the forementioned method in deeppaint3d or bodypaint using the projection option is PERFECT. No texture stretching or problems of the like. Localized resolution control. I'm sure there's more too!
Not perfect form computational point of view! I eman ist not efficent. And as siad theres a lot of other cosiderations except getting the pixels where you want them. YOu rarely actualy have any nee dfor this.
Still its not anew technique ist been used for ages. And manly because it has so many bad cabveats its been dismissed atleast as many times.
Joojaa: This is not an "efficent" way to map in maya. This idea and method is not in any way considering efficency. This is also not a "new" way of mapping, I'm sure.
That said, this method does work, and works well in my tests so far. This method is also not automated in any way in maya. This is why I'm posting in this forum.
I've been mapping with maya and photoshop for a long time now and I do not find the process as intuitive or enjoyable as painting on the model in realtime. Mapping uvs is a pain for everyone.
I feel the need to say this now since you're obviously not understanding why someone would want to do what I'm proposing:
I do not care about the current limitations of computers. This method is fast enough now.
sure, its something theve been working on for the last few centuries or so in cartography, without any real succes.
anyway a start is to do this follows something ive used in the past and acylay had sittin on one shelf. Altough to be honest id just lay them all out on one map! noy multiple ones its just plain stupid.
{// project per face 2002 $faces=ls -sl -fl; for ($face in $faces) polyAutoProjection -ibd 1 -cm 0 -l 2 -sc 2 -o 1 -p 6 -ps 0.2 -ch 0 $face; }
I mean you could use vertex colors aswell...
Ps use layout uves to get them all nicely lined up on one texture (it wont affect your maps disortion in any way!)
the uvmap can act as a guise for procedural textures.