Texel density in movies

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

Thread Tools Search this Thread Display Modes
Old 04 April 2013   #1
Texel density in movies


I was wondering, does anybody know what the workflow is texel density on animations/movies like tintin iron man or transformers or whatever? I always have this problem when doing big camera zooms and was wondering how they do it in the big studios. I looked at this zoom in tintin on the boat scene where the camera moves from way back to real close up the wall. There was no tiling textures or whatever.

Are those trazillion K texture maps? Is it Ptex? Do they work on a shot to shot basis maybe? I sometimes see camera's flying over a road which has no tiling and is one big piece of geometry, how the hell do they do that?!

Many thanks for any help, suggestions, explenations and/or articles on this!!
Old 04 April 2013   #2
for feature film multi tile texuring is the way to go...
20 to 200 uv tiles in most cases...
a hero asset can have up to thousand 4k maps...

take a look at the mari presentation here at 43min...
there are some insides on the cyclops..

you could also download here the blacksmith asset and take a closer look...
Old 04 April 2013   #3
Originally Posted by oglu: for feature film multi tile texuring is the way to go...
20 to 200 uv tiles in most cases...
a hero asset can have up to thousand 4k maps...


Also, you can just talk about texture resolution. No texture painter I know uses the term texel density :-) We usually just talk about maps by their size, eg 4k or 8k, etc.
Old 04 April 2013   #4

I'll have a look at multi tile texturing! You've been very helpfull. I didn't know what to look for. I think that multi tile texturing will be pretty hard, you'll probably get trouble with seems in normal maps and stuff like that.

Thanks again.
Old 04 April 2013   #5
Well with 3D painting apps becoming the norm these days, things like texture seams are a non-issue. Applications like Mari paint seamlessly across meshes and are designed specifically to deal with multi-UV tile layouts.
Old 04 April 2013   #6
The stupid thing is that I've used this technique for a couple of years now. Sinds I started texturing in mudbox I layed out my UV's across multiple UV tiles..

For some reason I never thought of going as high as 40 tiles! Always kept it at three and mostly just for characters: head in one tile, body in the next tile. For some reason I never thought of dividing the head itself over multiple tiles like the blacksmith model.

It really stupid LOL
Old 04 April 2013   #7
On top of that, some places may also use multiple UV sets for certain things, for example a furry character may not need as detailed texture maps for most of it's body as it's covered in fur. So they may only use something like 14-20 tiles, but then have a second UV to allow for key areas like eyes nose and mouth, claws and any other hero area that isn't covered by fur.
Quote: "Until you do what you believe in, how do you know whether you believe in it or not?" -Leo Tolstoy
Kai Pedersen
Old 04 April 2013   #8
Thumbs down

error duplicate post
Quote: "Until you do what you believe in, how do you know whether you believe in it or not?" -Leo Tolstoy
Kai Pedersen

Last edited by LucentDreams : 05 May 2013 at 06:24 PM.
Old 05 May 2013   #9
Interesting discussion, one thing I was wondering about this, is why would you use several low rez textures instead of one big texture. For example the blacksmith head from mari use 1k textures on several tiles. Why not making less tiles and use one 8k per tile?
Old 05 May 2013   #10
If all I'm rendering is a close up of the front of the head, why do I want to load a massive 8k texture into memory that includes the neck and back of the head? It's a large scale way of doing what Mipmapping does which most feature quality renders do well, and allow permanent conversion which can save a little time compared to on the fly like most commercial renders will do. Mip maps are perfect square tiled textures, so even a 1k image is broken into many smaller tiles as well as multiple resolutions of the image.

Another is also on the painting and editing, working with a 1k image is always going to be faster than an 8k image in every application. even if I load in several tiles at once, say 8, that's 8k wide while not being 8k tall so I've still loaded in 1/8th of the data that the 8k would have to load to edit. Especially with a modern app like Mari, when you zoom in and out and edit textures its dynamically adjusting their preview resolutions. when you zoom way in on an area it will show it in all of it's highres glory because it can fit it all into graphics memory but there's no way it's going to fit a lot of layers or tiles for an 8k image, as a single layer/tile for 8k is going to be 256mb, on a 2GB card that's a max of 8 simultaneous not including the mesh and shading etc that also is on that cards memory, while a 1k is about 4mb of memory so it can easy load and unload 500 layers/tiles (simplified math)

I should add that 1K isn't a rule by any mean either, you can very well have a 2k, 4k, 8k or even 16k image if there's a good reason for it, It's a matter of finding a good balance, We often have 2k and 4k textures. One of the great things about this workflow is that if they decide later on to have a hero shot of his fingers, you can go and scale up those tiles and then refine, without affecting the rest of the model so having more small tiles means you can be more selective in what you need to scale up.

I mean this is sort of what makes Ptex so effective, each polygon has it's own texel grid so you can manage resolution on every single polygon. The disadvantage to doing that ourselfs wold be the large number of textures to manage, which is why Ptex makes a container that then stores all the Ptex and their coordinates in a single file instead of a UV and multiple images.
Quote: "Until you do what you believe in, how do you know whether you believe in it or not?" -Leo Tolstoy
Kai Pedersen

Last edited by LucentDreams : 05 May 2013 at 06:22 PM.
Old 05 May 2013   #11
Thanks for the answer, the explanation is really great and that totally make sense.
If you allow me an other question, why would you still use conventionnal uvlayout instead of Ptex if you are using many tiles?
The only advantage I see about using "regular" uvs is for editing your maps in photoshop or projecting your texture in uv space.
But if you are using multiple tiles the editing won't be that usefull right?
Old 05 May 2013   #12
Ptex has it's own challenges, the key one being that modifications to mesh after texturing is well underway are more difficult to use. You have to resort to re-projecting from one mesh to the other so that texture information on the polygons that may have been added or removed gets projected into the new topology. You also still end up needing UV's for certain kinds of tools later on perhaps in FX, or for certain deformers etc. UV space is a commonly used thing for more than just texturing, so while the idea that ptex eliminates the need for UV's is a nice one, in more complex pipelines that isn't always the case.

Finally there's also simply the time to test and implement ptex into an existing pipeline. I do believe PTex is a very significant technology, but it's still going to need time to mature and be adopted and that time has a lot less to do with the artist and more to do with the tech/pipeline side of things.
Quote: "Until you do what you believe in, how do you know whether you believe in it or not?" -Leo Tolstoy
Kai Pedersen
Old 05 May 2013   #13
Thank you, I have to take a look at the other use of uvs in the pipeline, that's interesting to know.
Old 05 May 2013   #14
ptex is really great for everything you want to bake... displacement, ao or custom data...
if the topo changes just create new ptex and rebake...

or for storing data for hair creation... all kind of special maps...
Old 05 May 2013   #15
Texel density needs to be 2x the resolution of the final render resolution. To calculate, create a black and white checkerboard texture with each alternating pixel. Create at 512k resolution and assign as a surface shader without any shading module. In your render settings make sure all texture filtering and aliasing are turned off. Then move your camera to the closest you will ever get to the mesh. If you see a checkerboard or moire pattern, the texel density is to low. Double the resolution till you get only 50% Grey.
: my website :
Thread Closed share thread

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Society of Digital Artists

Powered by vBulletin
Copyright ©2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump

All times are GMT. The time now is 02:18 PM.

Powered by vBulletin
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.