I wonder what would be the most correct way to comp reflection over a glass surface line a windshield or a telephone screen.
Recently, I had to do that as the final touch of a comp and after creating a reflection pass on a black BG, I used Screen mode to comp it.
But i was not happy with the result.
The bright parts of the reflection are basically added to what’s already on the screen and that creates very bright parts in the final image. That also makes some parts brighter than the reflection pass which I don’t think looks realistic.
In my mind, bright parts of the reflection should be opaque and dark parts should be transparent.
But if I use the luminance of the reflection pass as a matte to comp it in blend mode, it creates ugly semi transparent areas…
Hope I’m clear enough…
Screen or add should be the way to go. You can use a color corrector node on the footage before comping it lessen the levels on the highlights and/or darken the blacks.
Depending on what kind of footage you’re working with, you might need to clamp your footage at some point so you don’t end up with weird values though
In my opinion, the correct way is the way that makes it look good and gets the shot approved!
I LOVE screen comps…I mean…scromps (you’r welcome, @vfxjamie). They are usually pretty straight forward, but I see so many online that are just janky. I have uploaded a scromps Batch Setup in the Logik Portal if you’d like to take a look at a basic scromp build. And, why the heck not, here’s a little live example as to how I do scromps. (Right click and download for 4k viewing)
Yes, nice. Simply didn’t think about lowering the highlights of my background plate before comp. will try that.
Actually, I mostly need background added to reflection to not be brighter than reflection pass.
I think then I should lower highlights of the background plate using the reflection pass as a matte.(lower in brighter parts of reflection but not in dark parts)
I’ve heard that ‘Screen mode’ was invented by John Knoll at ILM to simulate what happens when you optically add one layer to another, but for sRGB images in Photoshop. ( I could be wrong)
So the correct way to add reflections is to literally ADD the image but in Linear color space because that’s what happens in real life.
If you ever comped additive glows in Linear color space, and then converted your comp to sRGB or rec709, you’ll notice that it just looks like a screen. But what’s funny is that even though this is “correct”, I would rather comp glows using add mode in rec709 because it looks better (more blown out and stylized). To get the same look in Linear, you have to crank up your glows past 100% white. I saw a tutorial once where they showed the correct way to comp light sabers is to use primary colors but with > 100% RGB values so that when the exposure falls off, you get vibrant colors on the edges. Just like our sun is yellow but looks white because it’s so bright.
And alternatively you can simulate a neutral density filter on a camera or exposure change by subtracting a transparent black frame in Linear footage. This is useful to get an optical fade or simulate a tinted car window. The super whites will not turn grey or muddy as quickly as it does in sRGB/rec709.
But for a screen comp, I just tweak the reflection by eye.
I remember someone posted on Logik facebook a matchbox or batch setup that simulates a logarithmic screen that Nuke artists use that make realistic looking reflections. I did a search and I can’t find it. Maybe someone knows what I’m talking about???
There’s also the element of how reflections relate to 3d space as well, in regards to a camera move beyond a nodal pan. There’s all sorts of science and e=mc squares and whatnot involved, but I find I can think about, it in the most simple terms, as what’s reflecting in the screen is going to be tracking at the same z depth as if it was behind the screen. So if I do a 3d track and I have a locator that is on the window 5 feet behind from the screen, and I want to reflect a wall that’s five feet in front of the screen, that same locator is going to work perfectly fine.
My personal go to, is to use the Average blend mode in Flame to comp the BG plate back over the screen insert and then a hold out matte for the screen to comp the picture back over it. Then I’ll color correct the artwork up or down to a level that the client likes. You can also use a separate cc to adjust the AVG layer to your tastes.
Again, like every one says, its visual preference thing. What clients want isn’t always what looks realistic unfortunately.
Off topic - I had a client that wanted all the artwork to stay in focus so it was readable in the screen, regardless of the fact that the plate was going in and out of focus. They also wanted the artwork much brighter than the plate because they wanted it to stand out. Looked like shite, but after revision 5, I caved and said f**k it. If that’s what they want, then that’s what they want.
I have a custom hotkey for toggling the blend modes, for surfaces and geometry in action and also for the comp node in batch.
I keep clicking it and watch the result and stop when I like it for further adjustments.
Sometimes I click too fast and go on clicking and wait for the next round. Like https://youtu.be/LPguqGyV4OU
Here’s the thing: my experience of this is making adverts. Adverts which are there to sell a product or service a company provides. If the focus of the advert is centred on an app or website and the client needs this message to get across and has been adamant about this since the PPM, I don’t understand why a creative director, dop or director generally decide to film the device/laptop with a big white highlight across it and then encourage the editor to use the take that obscures the whole point of the advert? And yet here we are recounting this situation ad nauseum trading tricks to get around it. And on that subject, best trick I’ve found is get rid of the reflection off the entire phone completely even outside the screen area. Do your comp, make it look good and add back the reflection to taste.
Also Aaaaaaaaaaaaaaaaaaaandy Dill’s stabilisation trick for screens could be quite useful here.