Just for fun

[Paul Oblomov]

Well, any progress ?
Would love to save calculated solution and reuse it, so that it still render in that "unbiased" way

[vlado]

Well, any progress ?
Would love to save calculated solution and reuse it, so that it still render in that "unbiased" way

One way to do this is to split the rendering into elements; the caustics will go in their own element, so you can composite them later on.

With this method, caustics are splatted directly on the screen, so it a very view-dependent method and it is not possible to reuse them in the same way as a normal caustics photon map.

Best regards,
Vlado

[Paul Oblomov]

Hm. Can then interpolation of saved map be slightly noisier, but more detailed ? Like the picture I get when turning this experimental option on ?

[jacksc02]

have you done any images with this paul?

[vlado]

Hm. Can then interpolation of saved map be slightly noisier, but more detailed ? Like the picture I get when turning this experimental option on ?

Yes, of course. To do this, set the Max. photons option to 0, and set the search distance to match approximately the size of a pixel or two. However, to get a smooth coverage you will need many, many, many photons. It is how I did this picture, for example, where all lighting is done with photons:



In general, this looks like the most robust method for taking into account all light transport paths - indeed, the only one that can do the caustics-in-mirrors effects efficiently, so we'll be doing more research in this direction...

Best regards,
Vlado

[jacksc02]

looking good vlado

[studioDIM]

One thing i noticed is that it takes in account the un-smoothed normal.
In the case of a simple default sphere, i get some really interesting patterns around the main burn, due to the tracer correctly bending light rays per poly...
An option to take into account smoothed normals may be helpful, if it's at all possible with this method.

Lele

[Paul Oblomov]

vlado
Ok. Will try that

jacksc02
Well, all my tests so far >

[studioDIM]

Nice one, Paul!

I was wondering how this is tied to the PPT, and lol...
It seems if i have caustics on (using this method) and turn ppt on as a second pass, the search distance in the caustics tab somewhat drives how the ppt sees the pre-traced caustics...
Being search dist. an absolute value, it's like trying to set the ppt sample size in world units...

So funny to toy with...

Lele

[dellis]

Forget caustics.....more Cow Bell!

[putyi]

Just thought I bump this thread.

Specular reflected and refracted light always gave us a hard time, however they can add quite a lot to the overall realism, and feel of the image, especially in some special cases. Adding a max density control to the caustics photons too was a good step forward, but imo what is presented here in this thread *is* the thing. Having the photons traced directly on screen instead into a chahce is a good way to go in many situations, even if the current method can't support the refracted/reflected image of the caustics.

One thing, I can't understand, is the high memory usage. As far as I know, one advantage of a method like this over a chached, then interpolated one would be a much lover memory consumption. But I suppose, right now besides the directly rendered caustics, a cached map also being stored too, so besides trading in splotches for noise, there's no actual gain. Anyway, building light fixtures is still fun.

Regardless of all the silly unbiased hype around, I do think this has some nice potential. So, how are things, Vlado?

best regards

edit.: Just realized that the max density has no effect on the direct tracing, only on the stored map, so, I guess my memory issues are solved. Hehe...

[vlado]

So, how are things, Vlado?

They are coming along

Best regards,
Vlado

[putyi]

Great news! Perfect to make a monday jolly!

best regards

[pskdude]

lmao

i agree with Dellis more cowbells LOL !

[Buck]

I'd love to have super sharp caustics like that! I've done some 'caustic art' before, and would like to do it some more with those updated caustics!