ARRI Tech Talk: A DP's Analysis of the new REVEAL Color Science (subtitles EN,ES,FR,PT,한국어,日本語)

REVEAL Color Science is a completely new image processing chain used by ALEXA 35 internally and available for ARRIRAW from ALEXA 35, ALEXA LF, and ALEXA Mini LF in postproduction. In this Tech Talk accomplished cinematographer and Cinema Lens Specialist, Art Adams, walks you through several practical examples of the improvements you will discover within REVEAL, compared to the older color science.
Find out more about REVEAL Color Science: www.arri.com/e...

[@TuckerDowns]

Great video and guidance on using the new cameras, thanks!

[@nerdstalgia-1]

The reds in the sign in the background @14:50 seem to be clipping weird in the Reveal color science, probably some KZbin color conversion thingy. I think the original is fine as the clipping only appears after fully fading to the reveal version so the original shot seems to be fine.

[@javierhick4082]

Thanks! Very interesting

[@Owlbot]

Art is the GOAT!

[@WildShadowsZA]

Excellent

[@0800filmez]

I always felt that Ryan Johnson missed it out by shooting Digital back then. The Reds on Star Wars shot on film are quite different. And that changes everything.

[@bobh1782]

tell that to steve yedlin haha

[@Fedor_Dokuchaev_Color]

His DoP (Steve Yedlin) has created his own transform math to make digital look like film. His fellow DoP's could not distinguish between two pictures - one digital and one shot on film. I would suggest to read his article "On color science" where he explains his thoughts in more detail.

[@johnantohe226]

wooow!!!!👏👏👏

[@robertulrich3964]

@Art Adams Greetings Mr. Adams, can you clarify if the previous Arri camera color science rolled off saturation or compressed it non-linearly? I saw your older video on provideocoalition about the "film look and why there's two of them" and you stated "As exposure increased, colors got brighter but they didn’t get more saturated". Many thanks Sir!

[@art_adams_arri]

That's a good question. When I think "compressed," I think about the differences between the various color curves dropping to the point where there's effectively no room for them to move in relation to each other. This was a trick with some early HD cameras where the gamma curves effectively merged as they approached highlight clip. That was reasonably effective at preventing hues shifts in highlights but still resulted in blocky clipped whites. My understanding is that ARRI is doing some different things and the roll-off is a combination of display transform LUT and the way we designed our color space. When playing with Reveal in Resolve, I noticed something interesting. If I used a color managed or ACES workflow, the roll-off in highlight saturation was uneven. Yellow continued to saturate while other colors became brighter and less saturated. When I chose to work in LogC4, all hues rolled off at the same rate and the result was much more pleasant. There are ways to do that in ACES, but in our own color space it's dead easy. (We auditioned several color spaces with colorists before we settled on the one we use now, by the way.)

[@lilsouschef]

​@@art_adams_arri Curious - how would I be able to replicate a test to see what you saw with the yellows saturating while the other colors became brighter? I have downloaded ARRIRAW footage from the website, but I'm not as advanced as you but would like to observe this to better understand! Thank you for the video!

[@art_adams_arri]

@@lilsouschef What I did was I shot a color chart in ARRIRAW and then processed it in Resolve using several different color spaces. I then pushed the gain way, way up until colors started to clip. LogC3/LogC4 both did a great job of desaturating hues in a perceptually equal manner. Some of the others didn't, and yellow seemed the most problematic.

[@lilsouschef]

ah, thank you so much for this insight!@@art_adams_arri

[@TuckerDowns]

A bit misleading to just talk about blue being 8% of our brightness sensitivity @6:13, that would imply that the green patch is almost 10X brighter than the blue patch. That might be true in terms of luminance but it we talk about perceived lightness or perceived brightness, the HK effect has a huge influence on the perception of saturated colors like these, making them appear brighter than their luminance would imply.

[@art_adams_arri]

Thanks for that. Everything in terms of color is a bit of a rabbit hole, so I'm trying to keep things simpler for videos like this.

[@TuckerDowns]

@@art_adams_arri Makes perfect sense. :) It's a great video none the less! Just a color scientist being pedantic.

[@art_adams_arri]

@@TuckerDowns And I completely appreciate that. :) That's one of the better ways for people like me to learn.

[@nshea3286]

Isn't luminance the actual perceived brightness since it is photometric measure. If two colors are equally saturated wouldn't the HK effect be much less an influence on perceived brightness? I know that blues at the same power(radiance) as a Green or yellow would appear much dimmer because we have far fewer Blue receptors in our eyes than Green. The Blue light's power output(actual photons) would probably have to be ( watts not lux) 10x the Green to appear the same brightness

[@kedbear79]

A lot of this is just looking at 2 different LUTS? Who ever used the arri K1S1 logc to 709 LUT anyway? How may of these Reveal characteristics are coming from the debayer to awg4 logc4 vs the Reveal ODT's? In any case, you can take the old Alev3 sensor and via a custom transform / LUT make it behave like Reveal if you wished, eg getting those tail light reds to go orange at hi sat is no problem etc etc.

[@gafpagear]

Agreed Any bayer sensor uses a 3x3 matrix that calculation isn't improved and still the same. All they did was to apply a little bit vintage huesatdim lut, because a lot of dops had issues with hues passing a certain luminance threshold to become pastels etc. This os all very easy to map differently. When you can populate a y5 cube lut and feed a wide colorspave programs like light match by light illusion could render such a non destructive 65 cube log transform. Most in this video is just pure marketing bullshit. Colourscience is an odd word anyhow. As long as codecs are good and spectral sensitivity is good one can map the camera like ge wants. Though it's Easier and slightly better to do this in lineair. Though it might surprise how little the differences are if the math is right.

[@kedbear79]

@@gafpagear What gives me pause for thought is Arri state there is 'undisclosed' and non Matrix based math going from camera native to awg4 in the debayer. The math is no longer using a Matrix like Alev3 when it debayers to awg3. So this is interesting.

[@gafpagear]

@kedbear79 well its possible. Obviously so matrix is 100 010 001. In that case the huesatdim map which has a non lineair behaviour has to fix everything. But arri had made quite some false claims in the past about colors . And art Adams is not a colorscientist. A matrix is a linear formula at its best its calculated through delta00 which us perfect. But a huesatdim map can track the perceptual changes in high saturation and luminance , its something I prefer a camera not to have because it's harder to reverse than to add.

[@kedbear79]

@@gafpagear I think the issue in the old colour science was it was 100% Matrix based ie a Matrix in the debayer to awg3 and then a Matrix in the K1S1 logc to 709 LUT. Because the Matrix was essentially saturating and is linear the high sat hit/clipped edge gamut and the result is eg magenta clipped tail lights. In the new colour science the debayer to awg4, whatever non linear process it is will be quite gentle, however the non linear treatment is also in the new ODT's and that's the critical part for controlling hi sat hi luma behaviour eg not clipping them. Clearly unlike the old colour science the new is non linear both in the debayer and in the ODT. However, I'm just curious re the debayer process because if you want to map a camera into awg4 eg via a CST I assume resolve is using a Matrix to do that and therefore won't be using the non linear undisclosed math that Arri use in their debayer / SDK for Alexa 35 / mini LF when shooting ARRIRAW.

[@art_adams_arri]

I'm not sure what you mean by "looking at 2 different LUTs". The differences that I show are not related to display LUTs, although those do have an effect on the images. The difference is the math that we use to translate sensor values into color. It's completely different from anything we've done before and is unique in terms of in-camera processing tools. The biggest question we get is, "How do you make a Mini LF match Alexa 35/Reveal on set?" If you can make a LUT that will do that, you can make a lot of money. :)

[@theshaktinath]

Any paper on this?

[@ARRIChannel]

Hey there, yes we have many white papers available on REVEAL and ALEXA 35 workflows available here: www.arri.com/en/learn-help/learn-help-camera-system/technical-downloads

[@theshaktinath]

@@ARRIChannel wow! Thank u, never thought u guys reply...thank u once again, love from Nepal...

[@VariTimo]

I would to see Arri implement a standard for HDR viewing across viewing platforms. Right know there is no way to truly make sure that the finished product will look the same across platforms. Part of that are the absolute values of HDR across screens with different capabilities. But we cannot rely on TV and monitor manufacturers to enforce that.

[@Jorge_Ambruster]

That's actually what Dolby Vision was created for. It's just not standardized yet

[@Jorge_Ambruster]

@@kemalettinsrt Agreed. It was supposed to be a color space standard substitute for Rec.709 but it has come to the same problem, there's no fixed way to do HDR and because of that every screen that does HDR has different luminance values and different colors. Dolby Vision, from what I heard from someone working there, was supposed to address this. It's supposed to be a container for the maximum color and dynamic range information plus it works smartly so that even if your screen has a smaller maximum brightness than mine, it will adapt it and shrink it to make it fit instead of doing what HDR in Rec.2020 does which is cutting it from the highlights

[@VariTimo]

@@Jorge_Ambruster I know a few movies that don’t look the way they’re supposed to even on Dolby Vision.

[@VariTimo]

@@kemalettinsrt Oh yeah HDR is definitely real dumb.

[@Jorge_Ambruster]

@@VariTimo Me too. As I said before it's just what is supposed to do. But we haven't achieved it. We have to eventually figure a system that gives you the same results no matter the luminance or the quality of your screen. Dolby Vision, HDR and all this stuff was supposed to be that but it isn't

[@MichaelWeizenfeld]

Excuse me Art Adams and Arri, but in cinema we DONT need a human-like color reproduction, we need enchanced harmonized color reproduction which both Kodak film stock and old Alexa did. With this new colors you put your new Alexa close to the same category as Red or Blackmagic or Sony in this aspect.