Author Topic: when will new caustics solver be done?  (Read 2596 times)

2020-07-23, 19:03:05

johnmarc

  • Active Users
  • **
  • Posts: 23
    • View Profile

2020-07-23, 19:26:38
Reply #1

TomG

  • Corona Team
  • Active Users
  • ****
  • Posts: 3387
    • View Profile
Nothing is ever "done", but iteratively gets better - the same will apply to caustics as to the rest of the render engine, where "GI bouncing" is never "done" as we find ways to make it better, and so on for all parts of the engine. As for when there will be the next iterative improvements, the roadmap is the source for that- https://trello.com/b/EfPE4kPx/corona-tentative-road-map-3ds-max

2020-08-04, 17:23:49
Reply #2

johnmarc

  • Active Users
  • **
  • Posts: 23
    • View Profile
Does rc1 have better caustic after exposing how bad caustics solver is or is that in rc2

2020-08-04, 17:29:58
Reply #3

TomG

  • Corona Team
  • Active Users
  • ****
  • Posts: 3387
    • View Profile
There are no changes to caustics in Corona Renderer 6 other than those noted in the changelog, see https://forum.corona-renderer.com/index.php?topic=26830

2020-08-04, 17:37:46
Reply #4

TomG

  • Corona Team
  • Active Users
  • ****
  • Posts: 3387
    • View Profile

2020-08-04, 18:15:46
Reply #5

johnmarc

  • Active Users
  • **
  • Posts: 23
    • View Profile
but your caustics solver is terrible even Maru and your whole community knows this

2020-08-04, 18:24:08
Reply #6

johnmarc

  • Active Users
  • **
  • Posts: 23
    • View Profile

2020-08-04, 18:26:44
Reply #7

johnmarc

  • Active Users
  • **
  • Posts: 23
    • View Profile

2020-08-04, 18:52:50
Reply #8

TomG

  • Corona Team
  • Active Users
  • ****
  • Posts: 3387
    • View Profile
We'll have to agree to disagree on it being terrible, and on how many people think that way (vs. how many are already putting it to use). Either way, improving something as inherently complex as a caustics solver is something that takes a long time, it is not a trivial task.

2020-08-05, 11:24:10
Reply #9

jrgby

  • Active Users
  • **
  • Posts: 66
    • View Profile
I'm happy to wait for incremental improvements to what has to be one of the most complicated parts of the engine.

+1 for the Corona teams patience with threads like this.

2020-08-05, 16:38:07
Reply #10

dj_buckley

  • Active Users
  • **
  • Posts: 332
    • View Profile
Haha don't bring me into this.  I was merely sharing my findings.  I'm yet to be shown a 'commissioned production scene' using Caustics that isn't an outdoor pool in bright sunlight but hey ho ;)

2020-08-05, 17:00:47
Reply #11

lolec

  • Active Users
  • **
  • Posts: 156
    • View Profile
A new solver was never promised.

Caustics are notoriously hard, all render engines struggle with this problem in some form or another. Many don't have it implemented at all. Knowing the time and effort it would take to create a new solver, I think the majority of the community would prefer that time to be spent on other features.

I hear you, caustics on pools is pretty much the only commercial use case for caustics right now. But the thing is, that is pretty much the only time is really needed. Would caustics add a tiny amount of realism in some scenes? yes.

Look at this examples of architecture photography, can you find many -if any- examples of scenes that absolutely need caustics?  I couldn't find examples that even show clearly visible caustics.

In your study https://forum.corona-renderer.com/index.php?topic=30007.msg172830#msg172830 you showed a few examples on how caustics are visible in everyday life*

Except, in everyday life, no one is running around with a water bottle or hose spraying water all over the floor.

Of course you would get caustics in those scenarios... but unless you specialize in a very niche market of scenes-that-were-just-sprayed-with-water I don't see the current caustics solver holding anyone back.


2020-08-05, 17:05:00
Reply #12

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 10295
  • Marcin
    • View Profile
That's actually a very important question and we are super curious about the answer:
In what specific scenarios would you like to use caustics in your scenes?

I can think of:
- pools ;)
- jewelry
- close-ups showing glassware with water/other liquids
- there are some "eye candy" effects like: sunlight reflected from windows on a building facade, sunlight reflections from cars passing through the street

2020-08-05, 17:11:01
Reply #13

lolec

  • Active Users
  • **
  • Posts: 156
    • View Profile
Also, bringing dj_buckley in a little further :) - sorry! - you said if the goal is ultimate realism, questioning the utility of caustics is odd.

I agree.

But it is clear Ultimate realism is not the goal of Corona Render.

From the beginning, Corona characterized for choosing an interesting set of compromises that brought us Speed, Ease of use and "very reasonable realism"

In other words, Corona chose to tone down the realism just a little bit and just in the right way so the engine was very fast and very easy to use and configure.

Can you do more realism? sure. at the expense of time.

Can it be even easier to use and even faster? Yes! at the expense of realism.

Corona is not about ultimate realism. It's about equilibrium.
« Last Edit: 2020-08-05, 17:14:55 by lolec »

2020-08-06, 18:52:59
Reply #14

johnmarc

  • Active Users
  • **
  • Posts: 23
    • View Profile
The caustics solver needs major work and overhaul and should be number one priority of corona team as it is what Seperates
a mediocre render engine from world class as caustics are complex indirect illumination that when done correctly is what makes photo realism come to life in a render and right now caustics only work in boring simple scenes with easy light paths