VFB+: Filmic tonemapping not that effective?

[-Ben-Battler-]

Hi filmic tonemapping and VFB+ users!

I have a question concerning highlight compression.

I find that when compressing highlights with filmic tonemapping it is less "effective" than when using Corona's HC. Indeed the highlights get desaturated pleasingly but I'm not quite able to grasp how it works in detail. Sometimes parts of the highlights get brighter, other parts get darker. Same behaviour happens with Rich Shadows.

I am now having an interior scene and I want to create a Pano from it. But the outside is blown out even though there are big windows and Compress Highlights is set to a high value.
Here's the pano.

Can you give me some tipps how you approach highlight compression with VFB+? Decrease Exposure and then brighten the dark colors and then re-adding contrast? Not really, right..?

I would be glad for your feedback.

Regards,
Ben

[Juraj]

Looks like material issue. How does that white paint look ?

[-Ben-Battler-]

Looks like material issue. How does that white paint look ?

Dead simple actually.. Diffuse Level is 0.7.

Edit: Forgot half of the image.

Here the material's data:

Diff = 255, Level = 0.7
Refl Col = 255, Level = 0.3
Gloss = 0.3
Just a bump map.

[Juraj]

And how does the scene look like ? Can you make axonometry screenshot from the outside left part ?

This is definitely user-setup mistake, because the whites in interior are nicely tonemapped, but the transition is very harsch. What is 3D and what not ? How was it composited ?

With that said, even in correct setup (full environment, physically correct materials), you would have to use artificial light to make such transition look real. No amount of tonemapping in the world can equilize the exposure difference of outdoor and indoor to such level that interior is fully bright.

With that said x2, filmic is "slightly" weaker (it doesn't go as far as 99 in Reinhard for example), but I never had to use it. I use 0.9 at most (and with Reinhard 8).

[-Ben-Battler-]

And how does the scene look like ? Can you make axonometry screenshot from the outside left part ?

See the 2 Viewport screenshots. Everything is 3D, I didn't use any backplates.
Maybe I just found the culprit. As you see, the window material's refraction has a RGB value of 245. So less light enters the interior. I will fix that and try again.

With that said, even in correct setup (full environment, physically correct materials), you would have to use artificial light to make such transition look real. No amount of tonemapping in the world can equilize the exposure difference of outdoor and indoor to such level that interior is fully bright.

Okay. So if your client asks to "make the outside darker" then you would use artificial lighting in the interior to brighten it up?

With that said x2, filmic is "slightly" weaker (it doesn't go as far as 99 in Reinhard for example), but I never had to use it. I use 0.9 at most (and with Reinhard 8).

Thanks for that information.

[Rotem]

What is the pixel color value of the outside areas with and without highlight compression?

[Juraj]

Why is the outside floor so bright ? It's also white paint ?

There is nothing wrong with reflective glass. That's a correct glass.

You can use "3D" tricks if you don't want artificial light. Actually some of them are real-world mimicking. Outside white paint, is not as bright as interior paint. Use 150-160RGB for outdoor white paint, use 190-210 for interior.

When clients asks me to make outside darker I explain why I won't do it :- ). And then I won't do it. But I don't have issue like this in outdoor/indoor scenes because I set them up correctly.

[-Ben-Battler-]

What is the pixel color value of the outside areas with and without highlight compression?

Hi Rotem

Since I can't be sure to sample the same pixel twice it won't be accurate (maybe there's a way to do this properly?). But in the sunny areas the average pixel value decreases from 4.0 to 3.0 when applying the tonemapping.
In a GI lit area (ceiling of the terrasse) it stays at 1.6. Maybe it decreases a bit as well, I can't tell.

From that point it seems to compress the highlights more or less as expected. It just somehow feels to me happening in a less linear manner than Corona's HC.

Why is the outside floor so bright ? It's also white paint ?

It's a tile material with an average value of 0.4. allthough I've seen that it's mapped incorrectly. But it shouldn't matter in this case.

There is nothing wrong with reflective glass. That's a correct glass.

Yes but it's not the reflection value that is 245, it's the refraction value.

Thanks for your inputs.

I will make some changes and report back.

[Juraj]

Ah I see, didn't read that correctly.

[-Ben-Battler-]

Okay, setting full refraction didn't give much better results, I expected otherwise.
Seems like it's just a very contrasty situation as Juraj said. I'm now having a joyride with his mentioned method of using different RGB values inside and outside and it seems to work that way.

Thanks for the help, it is appreciated.

[-Ben-Battler-]

I rerendered the pano, you can check it again here if you want.

Outside wall material has now Level 0.55
Inside wall material has now level 0.9

So it's a huge difference. But it does the trick, there is not a big difference visible.

Optical question:

The IOR of glass is something like 1.5. That means that only ~5% of the light coming from outside enters the inside if it hits it perpendicular?
I mean, that would explain a lot!

[Rotem]

Pano looks quite good.

The IOR of glass is something like 1.5. That means that only ~5% of the light coming from outside enters the inside if it hits it perpendicular?

How did you figure that? I was under the impression that IOR affects the direction of refracted light, not its intensity.

[-Ben-Battler-]

How did you figure that? I was under the impression that IOR affects the direction of refracted light, not its intensity.

Thanks!
Right now I am talking about the reflectiveness, not refraction. So the amount of light that gets reflected back and does not enter the room is about 95% (because only 5% get through)? I am not sure about this but theoretically something like that should happen, shouldn't it?

[Juraj]

Where did you take that "5perc" number from ?

1.52 IOR means 0.04 frontal specular reflectivity 100perc. at grazing, and this is disregarding the refraction.

[-Ben-Battler-]

Where did you take that "5perc" number from ?

1.52 IOR means 0.04 frontal specular reflectivity 100perc. at grazing, and this is disregarding the refraction.

Urghh! I inverted the meaning. Yes, 5% reflectivity, not 95%... Then there's no big deal.