The other thing I might do, for a close-up, would be to increase the subdivision on the plate and spoon. That's in the object properties, not in the material (which looks fine to me). The edges aren't quite round currently and subdivision might fix that.

I'm not sure how useful that is, since it's all Firefly. He's using the Atmosphere but I've got Cycles volumes. He's adjusting the Light nodes but I've got an HDRI dome. It's very good, I'm just not sure how it applies to Superfly.

Personally, if you want a look like the original, I would reduce the Roughness quite a bit. Maybe even to 0-0.05. That will give it some surface shininess and gloss. The default 0.5 setting is a reasonable catch-all but lower values make things shiny and glossy, higher values make things matte.

Those are showing right at the edges of the texture. The face texture map ends right in front of the ears like that. And you say this only affects some textures. I think the texture map just isn't painted all the way to the edge. The solution would be to extend the edges of the face texture, like hborre said.

The lines we used to get from not picking unimesh were cleaner than this, I don't think it's that. (I think it's still a good idea to pick unimesh skinning, I just don't think that's the cause of this specific issue.)

Also added more kelp this time, and pushed the volumetric box a little further from the camera. And increased the brightness of the sun light but switched it to inverse linear attenuation, which helps with the surface vs deep lighting.

It's not flat planes, the stipes are more cylindrical and the bulbs are round. And this isn't flat planes with a transmap for leaves. But that said, the blades are pretty thin.

I turned off SSS and it actually gets a little more colorful, which is not what I expected. But an improvement.

The short answer is that you want Displacement.

Displacement expects as input a height map indicating how much each point should be raised or lowered. https://docs.blender.org/manual/en/latest/render/shader_nodes/vector/displacement.html

VectorDisplacement expects a map of the 3-dimensional vectors pointing out the direction and distance each point of the surface should be moved. This is represented as color data. So instead of a height map, it should be a vector map created specifically for that purpose. https://docs.blender.org/manual/en/latest/render/shader_nodes/vector/vector_displacement.html

If you use a height map as input for VectorDisplacement, sometimes you will luck out and things will point the right direction, other times it will point the wrong direction. Largely random.

Do you need to pose them? Fab.com (was Sketchfab) has a lot of 3D models some of which are high quality, just not as Poser characters/props. Usually can be used as fbx. Quality varies but some look good. https://www.fab.com/category/3d-model/characters-creatures--animals  Prices tend to be high though.

drages posted at 3:06 AM Wed, 4 December 2024 - #4491892

Wow, this is great source! Do you have any idea that I can change the distance for this behavior or any kind of stats? I am searching that where is the mip map poser uses? 

I played around with it a bit. Thinking maybe the textures needed to be reloaded to update to 16K or something, but same result as you.

What is the reason you don't want to just turn off mip maps entirely? Since we know that is the cause. For the scene shown above, I don't think the texel swimming effect is likely to be relevant. In principle you could cheat and fake the benefit of mip maps by using lower resolution textures for objects that are far away (as long as they aren't moving from close to far in the same animation).

nerd posted at 2:47 PM Sat, 7 December 2024 - #4491964

Yes the concept of a "Meta-node" could be really useful. How to implement it so it actually makes things easier ... that's the hard part. It gets ridiculously complex really fast.

You could also set them up like getting values from a compound node. In that case you could just essentially have one universal compound node.

In principle you could also have a global image map, but since there's already a function to replace all maps in the object/scene, it's not necessary.

An AI detects your tumor that any doctor would have missed and that then raises moral questions?

We're discussing generative AI. That's a complete non sequitur.

If they can get into Daz, then https://www.renderosity.com/marketplace/products/152558/duf-pose-converter-for-poser-12 would be able to take the additional step into Poser.

Did you already try the BVH import that is available under File/Import? If that's what you're saying is not working, maybe that is a solvable problem with a little more detail.

This seems like a pretty good explanation of why mip maps are used: https://gdbooks.gitbooks.io/legacyopengl/content/Chapter7/Mip.html

It also becomes apparent why there is a line beyond which the texture would become blurry. Because you have gone past a certain distance from the camera and therefore transitioned to a different mip map.

Of course since it is meant to solve a problem with things in perspective, you might not want it if the things in your image are aligned that way.

Can you just double the render dimensions and downscale those output files to get the detail you need?

Even ignoring the moral issues, AI would mean losing consistency which is a big problem if you are trying to do multiple renders of the same people in the same space. For example for animation.

That said, if you want to do that there is Ken's script already.