Condensation Render Inconsistencies

Condensation Render Inconsistencies

Hi All.

I opened a project from last year (some beer cans with condensation) that I need to modify with new artwork. I found that rendering the same file in the latest version of Cheetah3D (C3D 7.2.1) gave significantly different results.
I've told my client the next set of cans would be exactly the same and I'm hoping to achieve it in the latest C3D.

If you look at the attached image, C3D v7.0 beta 26 with Shadow type set to 'Raytrace' gives me the results I need to replicate for my new project.

C3D v7.0 beta 27 or beta 28 changed something and I can no longer get the results I need. I've tried everything I can think of, including different Ray Depth settings and different Render Tag settings.

Is this a bug? I'm unsure, but there needs be a way to render water droplets.

I've linked to a simplified file to see if anyone can help. >> https://www.dropbox.com/s/roscyzus0acqxda/CanCondensationTest.jas.zip?dl=0

Thanks in advance!
--shift studio.

IIRC, in beta 27 Martin dropped refractive caustics from Falcon and introduced a new render engine based on transparent shadows instead.

That means of course that from then on all refractive objects would render differently.

I keep beta 25 for that reason and if you have beta 26 still available I'd recommend to use it for your project.

If you'd need some of the newer modeling features you can always model in the current version, save and open in an earlier beta for rendering.

Hi Misoversaturated.

I suspected it had to do with that kind of thing... I read all the release notes, but sometimes glossed over things where implications weren't immediately apparent to me.

I have kept several of the betas, including 7.0 b26 - so I probably will end up using that work-around as you suggest for now.

Rendering in old versions isn't a perfect solution however - as time goes by there could come a time where the old version won't run anymore or I need some more advanced feature that won't render in the older version.

With that in mind, I'm wondering if anyone can help with a solution using C3D v7.2.1?

As an aside, I always thought caustics were on the outside of the transparent object. Maybe inside is refractive and outside is reflective? If thats the case, sign me up as a supporter of getting refractive caustics back into Falcon

Thanks in advance if there are any other insights to be shared.
--shift studio.

update - feels like I'm getting something close to usable.
note: Trans blur above '0' in the Water material seems to really darken the droplets.

I'll report back if I find settings I like
--shift studio.

Shift Studio said:

Rendering in old versions isn't a perfect solution however - as time goes by there could come a time where the old version won't run anymore or I need some more advanced feature that won't render in the older version.

With that in mind, I'm wondering if anyone can help with a solution using C3D v7.2.1?

Click to expand...

Yes, that's my problem too, that the awesome new stuff like soft shadow IES lights cannot be combined with caustics.
I usually end up using Blender/Cycles for rendering because it has all the features and is way faster because of internal denoising.

For a solution inside C3D 7.2, if the matter is only the front droplets appearing too dark,
there is a trick involving a transparency value >1 with the dielectric shader.
The shading is apparently multiplied with this value, so higher means brighter (kinda emissive effect).
See attached example:

Aside: the switch away from refractive caustics sucked a huge amount of my enthusiasm for falcon away. It seems to me it turned refractive objects in falcon renders from being a huge positive to a huge negative. Now, there were problems in some cases but the benefits were huge. I wish we could have them back, at least as an option.

Thanks Missoversaturated and Podperson.

@MO I tried your tip, I still couldn't get quite what I wanted but can see that that may be useful. . Thanks for sharing!

I'm doing my project in 7.0b26 because I don't have time for anymore trial and error.
One day, if I ever get it right in a current C3D I'll share the results.

Thanks again.
--shift studio.

Multiple versions

I use Snow Leopard and Yosemite. I still prefer a skeuomorphic interface to a flat one, and my Photoshop CS2 doesn't work on Yosemite. Most of my newer software won't work on Snow Leopard. OS 9 Super Paint still has a beautiful effect I can't duplicate elsewhere, so I have that also. (Four Macs total.)

I retained C3D beta 25 for caustic refractions and 7.1.1 for Snow Leopard. 7.2.1 won't run on Snow Leopard, so it's on Yosemite. I don't know how many Mac users share my quirky approach to OS versions, but it's not unusual for software to be available in more than the most recent version as an option for older systems.

It seems to me that caustics aren't the problem here.

The underlying material has the "Fresnel" checkbox set. I think that's messing with the internal reflections of the droplets.

I also wonder if the droplets are too uniform in their height. Because they're all kind of the same size the cutoff where they stop catching light is too sudden.

If you could make the droplets more varying then the suddenness of the change in appearance of the droplets would be less disconcerting. I think that's the intention of the crumple modifier, but its frequency is 0 which is weird. Changing it to 1 or 5 produced slightly nicer falloff behavior in the droplets.

Just changing the crumple modifier's frequency and adding Misoversaturated's transparency trick produces pretty nice results for me (using 7.2.1).

Summary:

• I set the frequency of the crumple modifier to 5. (With a frequency of 0 it just has a fixed offset.)
• I set the transparency of the water to 1.02
• I toggled off "fresnel" on the can and made it somewhat reflective (with a blur of 0.02).
• Rendered in 7.2.1

podperson said:

It seems to me that caustics aren't the problem here.

The underlying material has the "Fresnel" checkbox set. I think that's messing with the internal reflections of the droplets.

I also wonder if the droplets are too uniform in their height. Because they're all kind of the same size the cutoff where they stop catching light is too sudden.

If you could make the droplets more varying then the suddenness of the change in appearance of the droplets would be less disconcerting. I think that's the intention of the crumple modifier, but its frequency is 0 which is weird. Changing it to 1 or 5 produced slightly nicer falloff behavior in the droplets.

Just changing the crumple modifier's frequency and adding Misoversaturated's transparency trick produces pretty nice results for me (using 7.2.1).

Summary:

• I set the frequency of the crumple modifier to 5. (With a frequency of 0 it just has a fixed offset.)
• I set the transparency of the water to 1.02
• I toggled off "fresnel" on the can and made it somewhat reflective (with a blur of 0.02).
• Rendered in 7.2.1

Click to expand...

That looks great and I will check it out tomorrow.
The reason for the crumple with frequency set it 0, was to apply a normal move of 0.0001 to the can. That is then subtracted via Boolean from the spherical droplet clumps. This makes the droplet clumps non-intersecting with the can. Make sense? Is that logic sound?
Thanks for the insights!

—shift studio.

Ah! Using Crumple to set a fixed normal offset is an interesting idea but doesn't seem necessary. I like the effect of using it to vary the offset of the droplets.

podperson said:

Summary:

• I set the frequency of the crumple modifier to 5. (With a frequency of 0 it just has a fixed offset.)
• I set the transparency of the water to 1.02
• I toggled off "fresnel" on the can and made it somewhat reflective (with a blur of 0.02).
• Rendered in 7.2.1

Click to expand...

Great result, thanks for instructions!

The fresnel settings in my experience can cause all kinds of weird problems and I never really found out why.
It's unfortunate that the fresnel property is just an on/off setting in the material properties panel instead of being accessible by nodes but then again Cheetah's organisation of properties being sorted between nodes, shader properties and material tag is generally very unfortunate (compared to Blender, where you can access everything and mix as you like even without OSL).

Shift Studio said:

The reason for the crumple with frequency set it 0, was to apply a normal move of 0.0001 to the can

Click to expand...

Thanks for sharing that trick!