Call for Ideas - New LuxCoreRender Features post-v2.10

[Juhis]

Replace Embree3 by Embree4

Perhaps this would also give motivation to look into deformation motionblur.
forward.

I agree: deformation motionblur would be very welcome feature.
Alembic files can deliver velocity information in blender these days and is supported by cycles so i think there must be some way to deliver velocity to luxcore also.

The other thing that i am missing is caustics that would see trough transparent objects (same with reflections - so refraction and reflections doesn't support light tracing - in blender it works) - maybe intel's path guiding technique would be one solution because blender is using that one and it works.
I mean this: see the images (the other one has glass feature on both objects / refraction: no caustics.. and the other screenshot has two objects where other one has glass shader and other one has metal shader and there is no caustics on reflection). Damn I am bad explaining things, i hope screenshots are clear enough to explain "problem"

But i think deformation motion blur is more important feature.

[CodeHD]

The other thing that i am missing is caustics that would see trough transparent objects (same with reflections - so refraction and reflections doesn't support light tracing - in blender it works) - maybe intel's path guiding technique would be one solution because blender is using that one and it works.
I mean this: see the images (the other one has glass feature on both objects / refraction: no caustics.. and the other screenshot has two objects where other one has glass shader and other one has metal shader and there is no caustics on reflection). Damn I am bad explaining things, i hope screenshots are clear enough to explain "problem"

FYI: Path guiding was already on the list, deformation motion blur has now been added.

About the issue you describe: This is the good-old topic of "SDS" paths.
For the specific scene you are showing, LuxCore already has the "CausticsCache" that can deal with this.
Of course this may not be sufficient for certain more complex scenes, but how much Path guiding will solve remains to be seen, I will say.
There is also the BiDirVM engine that s not exposed in BlendLuxCore. It may also be suitable for some cases. Further testing required.

[epilectrolytics]

LuxCore has several unique render methods like light tracing and Metropolis sampling that should be utilised in more ways.

Currently there is one hybrid render engine that does light tracing for caustics only on CPU and path tracing independently on the GPU and then adds both results.

In earlier versions Metropolis sampling was available on GPU but later removed (from Blender) because it works not effectively there.
Metropolis does big mutations to find efficient light transporting paths and then small mutations to find more in the area.
It should be possible to do big mutations on the CPU and then continue with the small mutations on the GPU (maybe vice versa depending on the scene).
That would be a hybrid renderer with kind of path guiding, using elements already built in LuxCore.

If that works, a next step could be to do light tracing (Metro) on the CPU and path tracing (Sobol/Random) on the GPU and combine it in a hybrid BiDir engine.
Other than the existing backward/forward engine this would connect the rays like in BiDirCPU.

It is quite possible that these ideas are not feasible (otherwise Dade might have already implemented them) but maybe it is worth to explore new combinations of already existing features in a future experimental branch.

[nostalphia]

Greetings =3,
Anisotropic rotation，
Adaptive subdivision，
float curve,
trigonometric

[CodeHD]

Anisotropic rotation，
Adaptive subdivision，
float curve,
trigonometric

Hello,
some quick feedback:

Anisottropic rotation: May need clarification, see the other thread you posted in.

Adaptive subdivision: Already on the list.

float curve: This is clear.

"triginometric": This needs explanation, what trigonomettric functions, and for which purposes?

[Egert_Kanep]

Perhaps when talking about anisotropic rotation and ways to achieve that we could take a bit wider look at the way texture sampling works.

I like when uv's can freely be edited in shader network. In luxcore we have the uvtest node, but this output color data and cannot be used as uv's anymore.


If there is a way to freely manipulate uv's like we do any other texture and use manipulated uv's for texture sampler it could open up many possibilities.
If we were to take it a step further same could be ideal for position data. Most render engines have this option, but I am not sure how uv's and position is evaluated in LuxCore, perhaps the core does not allow to modify this data easily on the fly.

But I think in long term there is not much point perhaps to solve things related to one feature but rather expand the whole texture sampling workflow and it could at the same time solve all those rotation issues as well. Since currently anisotropy is locked to existing uv attribute, but as stated if there is a flowmap or some uv map that has been modified it could be used directly inside shader graph instead of relying on the geometry attributes only.

Maybe the distort node could serve as some sort of guide as this is doing uv manipulation, but is just limited to textures and not for overriding uv's for other calculations.