Anyway. I needed to re-write parts of maketarget and makeclothes, meanwhile the work is hopefully done.
https://github.com/makehumancommunity/community-plugins-maketarget
handles now different meshes, at least in the theory, but at the moment only that mh2bot is available.
It also can help to adjust helpers for all those people who forgot the helper, when they did targets.
https://github.com/makehumancommunity/community-plugins-makeclothes
Same goes for makeclothes, it now works with MakeHuman II (which is not that different) but also includes those parts of makeskin which create materials for MakeHuman II. No makeskin needed anymore. Without adding a material it will act like before, just create a dummy material.
The materials are reduced to glTF standard, except that litsphere stuff. MakeTarget is already explained in Hugo. Changed MakeClothes will follow.
After that we tried own cube-maps. And here the chaos started. When one compares all these fine tutorials they all use a natural environment (e.g. Khronos). Now Elv created an own with some texts on it. They came out mirrored. I flipped the cube-coordinates ... well nothing changed at all
Okay I need to flip the image inside. Then I recognized that my mirroring when presenting metal shows the wrong side of the cube-map. I added a minus sign somewhere
Furthermore I got a wonderful plugin to create cube maps, which created a bunch of extra scenes (what for?), made itself persistent and did more crashes to Blender than renders. The first time I deleted it and it was still there. Great. I mean I do that with rm on my disk. I even deleted __pycache__ I am not a Windows guy. But it was still there. I think it was added to the blend then.
... I took the original file from Elv, the old blend file with the camera inside was ruined. The bad thing about that plugin was, it inserted itself before the normal render. Well these plugins are exactly what nobody needs.
Since that was too dangerous for our future work I even wrote a new (simple) plugin to build the cubemaps. Doing an own render process is not that simple but do-able (it is running in parallel in the background). I will upload it to my own github account, I guess or pack it inside a special directory which is "delivered" anyway.
A cube map is simply a "quadratic" camera doing renders with a focal angle of 90 degrees in all directions. One should put it a approximately to 1m height in the middle of the scene or room.
When doing an room and you want to see the whole character but do not want to use a fish-eye double the size of the room, the character is always between the middle of the room and a wall. So if the room is 10x10m there are 5 meters left. It is a fake, it is not a room. Assets can only be put near the walls!
.
The standard is a clean photo-studio with our own textures now. It are jpeg-textures to keep the size in github small. The other new ones will be on own file-server, because they are bigger (especially as we will use png there).
This is the backdrop used in standard:
- a fashion pose in front of a studio backdrop
and this is the room elv made:
- perspectives of the cubemap
So next steps will be a lot of error corrections (e.g. bone rolls for the blender exporter are wrong I just noticed a few days ago), off-screen render only works on Linux (and there not always correctly) and finding the reason gets worse since AI has wonderful solutions in google for questions I never asked. If nobody had done this before with pyside6 and openGL then AI can only hallucinate.
Greetings
Punkduck
