So…all the cars and caustics are CG?
If yes, well…as experiment it seems good for advertising, quite realistic to be believable.
I’m pretty sure those are all photos and certainly not an experiment from the OP.
I worked on that campaign, the project can be found on my Behance profile. Some cars are CGI, but most of them are photos. The red M5 is photography only.
Just out of curiosity… is there a specific reason why for the same campaign the customer decided to put a few CGI shots together with the photos?
Uhmm.. I’m quite confused now. 
Who is the author of the images posted?
And if it’s a compositing of real photo and CGI, what is photo and what is CGI?
Why put some total photo beside some others that are mixed CGI/photo or full CGI?
Could the author (or authors) give more info?
I think it should be more useful for everyone.
None of the image posted is a render.
On the Behance page of kosso_olli you will find more images of the same campaign made by several people, among all the images only 4 are renders (1-23-24-26) all the rest are photos.
@vlado caustics study
Thanks, so only the car is CGI?
Can someone get this idiot (troll) of the forum.
@madcat117 this is your final warning - stop spamming the forum. One more thread like this and we’ll be forced to remove your posting rights.
And once again - if you really want to help - it will be very beneficial if you have an example of a renderer that meets your expectations.
Also, those car caustics seem doable in CG to me. Anyone think those results are impossible to get? Might be fiddly but could get close I think.
LOL…I have no other comment other than the fact that trolling these forums is NOT cool and can only end in one result.
These have been achieved since V-Ray had photon-mapped caustics, nearly two decades ago.
As to matching an *exact* look, by the pixel (f.e. to integrate with a plate that has other parts of the same caustics), i highly doubt that’s doable by anyone at all, no matter the tech, currently.
Besides, of course, manually painting the caustics in.
LOL…I have no other comment other than the fact that trolling these forums is NOT cool and can only end in one result.
While i personally understand the user’s frustration and motive, i can’t but agree on the fact that posting as he did (here, and on the octane forums, for size) without even mentioning the work isn’t his, is truly bad form.
It’s construable as naivete’, which is minor an annoyance, but it could also be construable as appropriation, which is no minor mishap.
I’d think the right thing to do would be to amend that post on the Octane forums as well, even if Olli’s been royally patient.
Exactly, this is what I was getting at. You could do a version that looks as good and believable, but not the same placement of highlights. If you had a clean plate, you could create a version, I think, that looks just as believable to a client.
I’ve done some test in the past and given the same conditions Vray caustics reproduce real caustics very well.
Here is an example inspired from this scientific test:
https://sciencing.com/happens-light-…m-8557530.html
Results are very close and I’m expecting that in general if geometry, shaders and lights are accurate you should be able to get accurate results in every scene.
The real issue is that to do that in a large scene, even if you tweak the photon map you will still need to use a massive number of photons, so huge render time and a huge RAM amount.
Posting images and video examples that Vray cannot produce is not trolling at all in fact it only continues to highlight your arrogance that Vray is some kind of incredible render and its not because the boys over at Octane are light years ahead of your progressive caustics solver with their GPU photon tracer kernel but its okay to think your cool and special right because your product manager therefore you have this illusion in your head that you have some kind of authority over me which you don’t as I will continue to make fake accounts as I please 8)
Here is a file for you to play with, built from Cosmos assets (with some small modifications to shaders to make them work for caustics.).
I attached the RGB and the Caustics RE, containing all -and only- the caustics traced in the scene.
As linked, the Max file is set up to use the old photon mapping, as it’s currently much quicker than the progressive one.
You can however switch to progressive and watch the caustics pattern emerge from pass 3 onwards.
On my old PC (AMD 1900X, 3.5Ghz locked freq.), the photon map takes around 3 and a half minutes to compute, and the rest of the rendering goes in a couple of minutes, so it remains a bit noisy still.
Progressive caustics would need a slightly different render settings setup, to minimise the speed loss due to the current state of the tech, but would produce sharper and more accurate caustics thanks to MLT guidance.
One suggestion before you get banned from Chaos and Otoy forums, life is short, you should not waste your time like this.