Personally I would render the 3d model out as raster myself (DKC style) and just use normal sprites but depends on what you are trying to do.
As long as you keep everything in Blender for example size wise to a full integer and then just render out at say 32x32 for example it should work fine.
I have done extensive tests as the current project I am working on will be doing just this.
You can also us something like Pixel Over on Steam to do this too… Which funnily enough is based on Godot. I am going to try and start using that as it also allows indexing of colour pallets etc, as well so its all in one place.
So if object is 2mx1mx1m in 3d , how does it translate to size of viewport( if I set 8x8) and render size image is (8x8) Does it average it or just not works ?
Not sure I understand what the problem is. The viewport will render whatever its camera sees on the resolution specified for the viewport size. If you set it to 8x8, the viewport will render into an 8x8 texture.
Usually with pixel art it’s best to stay at the intended resolution. Sizing up and down tends to introduce blur and unsharp edges. If you can do the post processing at the final resolution, I’d suggest trying that.
To introduce more colour shades on higher resolution,
post process - apply filter to picture to create from it pixels ( such filter is available in Krita with different pixel size ) .
For this I’m asking about shader if it can work within export tool rather then run scene .
And downsize it to make it realsize pixels instead of upscaled .