For reference i purchased a EChenry 3d model (quite new to this) and tried applying these all with node rangler as suggested by someone else elsewhere tho the result is definately not correct as seen in next image.
As there are 4 different textures for each channel (base colour, metalic, roughness emisive and normal etc) this material uses UDIM tiles.
This video should help you (to do it without node wrangler)
(Sorry I have no idea if node wrangler will do this properly but it is pretty simple to do manually)
ok tyvm, so i followed the blender part of the tutorial and it worked with placing the 1001 of each texture but since theres 4 in total for each how would that work exactly as they seem to be different textures? (unless im wrong lol)
hey there thank you, though that video is a older version of blender and i cant seem to find that window on the right side of that screenshot.
The window to the right of the attached image is in the UV Edit window.
The N key opens the corresponding option.
UDIMS are a way of splitting up textures so each file is not too big, The UV’s are laid out in tiles, if you look at the UV’s you will see that there are 4 “blocks” and only one of them is in the 0-1 range, the others are outside of that range.
Like this (there are five tiles in this one);
When you set the image node to use UDIM tiles
Blender maps those five images to form one large texture.
Note that when you select the image that ends in 001 in the image node and set it to use UDIMs the 001 disappears from the image name. That is because the image node will use all 4 (or more) images and map them out to the UV tiles that exist.
The reason for using this (instead of using just one huge texture mapped from 0-1) is that many game and render engines have a limit to texture sizes, with UDIMs you get around that limit by splitting the textures into tiles.
4x4k images would be one huge 16k image.
But once combined in blender, don’t you effectively now have a 16k image and need all that VRAM to be available?
(Asking because this question has been on my brain for a long time, and I can’t determine the answer.)
Yes, it is still a very large texture, but split into bits. It does not save VRAM, but imagine for example in Blender you limit your texture size to 2K, with Udims this would be 4x2K images (still fairly detailed) if it were one huge 16k image you would go down from 1x16k to 2k (and loose loads of detail). Many game engines etc have settings to limit (downscale) textures for performance.
Yes- the way Blender handles UDIMs doesn’t really optimize anything, but if you’re exporting to a game-engine, it will
Maybe the new texture cache will make it better (if I understand it correctly)
thats what N pulls up
Unless you do it manually or via another app (like Substance Painter), were you can specify a different output resolution per tile.
So, assuming one groups the UV’s logically (face on one tile, legs on another, etc), then the face image can be 4k (where more detail is wanted), while the legs can be 2k (where likely less detail is needed).
Right, I just mean in terms of draw calls and VRAM usage, a UDIM is the same in Blender, you won’t see those optimizations except in a game engine
- Import base_color.1001.png after creating a new Image Texture node
- Change the image setting to UDIM in the UV edit window
※ The base_color.1001.png image must be selected to view the option.
If the image is empty, you cannot see the option. - Get the UDIM image from here. (just select XX.1001.png image)
- When the UDIM image is loaded normally, the number of tiles is created automatically.
Wasn’t the original idea of UDIM’s to “simply” have one single material but different “areas” for UV’s on different parts (and also resolution in the image) to simple not have the UV’s overlap when using different materials… with the same settings anyway (expect different maps) ??
Some more something for better organizing this.
The optimazation according for example using mipmaps to reduce draw calls was allways something game engines are better with. And so they also does this according UDIMS…
( Yes i know… some other renderengines on- or offline are able to load only that mipmap resolution needed for the rendering and blender just loads everthing… but i always wondered when one have some location in the scene when you need all the biggest mipmaps then the VRAM may also not be enough… but that’s another topic or has some low probability ?? ).