Substance Painters UV viewport is 1X1 texturespace.

So when working with non square textures you need to stretch the UV to cover 1X1 space.

Due to this Painter works from a stretched UV and that will stretch the textures placed on it.

A good example on problems you can get is when your trying to add a checker pattern to a UV'd sleeve on a shirt, in a bent arm position. Its a complicated shape, with folds and with a bend in the middle. So not always possible to keep it orientated horizontal or vertical.

Because of this your not going to get the same clean checkerpattern all over the arm. It will be distorted, due to shape but also heavily distorted just because of the UV ratio.

Would be sooo golden if Substance support 1X2/2X1 ratio textures all the way. Viewport, generators and masks etc.

I'm working on a asset that has almost 40 udims. When I bake the utility maps needed for all those udims (some needing 8k) my file size increases to 20GB+ because all the maps are stored in the substance file.

I'd suggest an option in the preferences where one could choose a seperate location for those maps so every substance (and autosave) doesn't have to be so large.

This would save time on saves and autosaves (right now it's taking 3+ minutes on a 20+ GB file) and would also make me save the file more often, which reduces time and work lost when substances crashes.

I noticed that the workflow is much smoother when I segment parts of the model in advance to be their own unique layers.

Yet, I have no feature to combine the resulted PNG, because each area of the mesh itself still belongs to the same UV map.

I want an option that'll allow me to combine PNG'S, when they combine in this manner:
On each resulting image, the area external to the seen parts on the model are considered as "0", blank
And the areas seen on the model are considered as "1", filled with information.
When combining the PNG'S together, only the used parts will enter the final image... While everything ells (external information) will be ommited.

This will allow me to make a single PNG out of many different layers, and can make my workflow smoother.

Because each material brought outside of the software has it's own layer system, and you can hide parts you don't need to see at the moment.
It's a very useful function, and I want to work with it better.

It would be nice to visualize the settings in real time by producing rays that extrude from each vertex (or a portion of them when dealing with large vertex count, or even just a few you can personally select on the surface) along the surface of the object. It would also make the settings and normalization parameters easy to get an intuitive feel for if they were projected out of the surface using simple lines to display min max distance (etc...). Yes, some of these values are incredibly low and probably beyond what can be consistently rendered with floating point approximation, however you could also include a 'scale' tool that allows you to scale up or down values or even include a logarithmic option to scale values orders of magnitude different.

If your hard drive fills up, autosave will get stuck causing all work to be lost. It should be aware when the hard drive where autosaves are located is too full and not attempt to autosave if there is not enough space, and warn the user that there is not enough space to save the autosave file.

Or be able to cancel an autosave that is in progress.

Also, a setting to limit the number of autosaves for a project.

Also, something to not autosave when nothing has changed in the project.

I can't believe this happens! The feature that is supposed to save your butt, will ruin your life if your hard drive is too full to complete the autosave! Please fix this somehow!

For example, I can't texture my character in Substance Painter that way - nor Substance nor Unreal Engine support this! The only option - bake maps inside Blender!

What I mean, is that I just can't do smooth gradients with 4-5 colors, which have different positions, and especially if I want to place gradients precisely around model, or lots of gradients with masks, but with lots of transitions (visible-not visible-visible etc - on ONE mask).

Also, It's not a user-friendly to position gradient based on "Position" map. It often gives me wrong direction of gradient or is generally slow to make gradient I want.
Something done like in 2d app (like Affinity Designer) would be much better option! I mean placing visual gradient scale directly in 3d viewport!