Color Management 101

[Alexander Mejia]

I found that the artists at our studio were more inclined to calibrate their displays after this color management 101 video I put together. I'd love to get your guys professional feedback.

There is some NWS language and some video gamer jokes in there, but we work at a game studio and are use to this stuff.

Feel free to share this with your company if you found it useful.

 

[Justin McAleece]

Nice video! Thanks.

 

[Brad Allen]

Ha ha - this is pretty awesome Alex - nice job

Only thing I found distracting was your eyeline when reading the script - not everyone has a teleprompter on hand and at their disposal though (including me!)

Thanks for posting

 

[Alexander Mejia]

Ha ha - this is pretty awesome Alex - nice job

Only thing I found distracting was your eyeline when reading the script - not everyone has a teleprompter on hand and at their disposal though (including me!)

Thanks for posting

Yeah, I think after this we're going to get a teleprompter if we do more videos like this.

 

[Ryan Sauve]

Awesome video, great job! I didn't find the eyeline distracting but if you're looking for a cheap teleprompter, I recommend pad promoter with an iPad or Android tablet. That's what I use for my corporate stuff and it's been awesome.

 

[Mike Harrington]

pretty funny....

the eyeline thing....just push the camera back 15 feet if you can and put a display under the lens....
I have a prompter but still do this 50% of the time

 

[Noel R.]

Very good Alexander, I really wish there was a video like this available years ago. It would have saved me from scratching my head till I was digging into my brain. Granted there is still a lot of information to know when it comes to color management, this still makes a very good intro. With DaVinci Resolve Lite being free, and the recent price drop on Dreamcolor monitors, this seems like a very good time for folks to learn how to color time.

 

[Stacey Spears]

I enjoyed that. I would use CIE1976 over 1931. It's more common among calibrators in the field.

 

[Alexander Mejia]

I enjoyed that. I would use CIE1976 over 1931. It's more common among calibrators in the field.

I chose 1931 because when you google "sRGB" or "Adobe RGB" you're usually presented with a 1931. Although the basics of determining gamut are the same, it can be daunting when you're trying to find your own information off the bat when you're trying to find more info.

 

[Drew Baird]

Nice video. I posted it on Facebook and the response was almost instantly two people saying they know you. Small world.

 

[Alexander Mejia]

Cool. Who was it? You can PM me if you want.

 

[Gavin Greenwalt]

Your gamma description is mostly correct but the chart is off. 50% gray is in fact 20% gray and it's supposed to be 20% gray. That's why we hold up "18% gray cards" not "50% gray cards" for white balancing. CRTs used to make 18% gray less than 18% gray like 2% gray. So we would jack our 18% grays up to 50% gray so that the defect in the CRT would end up back at 18% gray. 50% gray isn't actually middle gray, it's exactly 1 stop below blown out white. Your graph though starts from linear color space (which it isn't) and goes to less than linear. It should start from the red (gamma corrected) line and map down to the linear line.

/nitpick

 

[Alexander Mejia]

Your gamma description is mostly correct but the chart is off. 50% gray is in fact 20% gray and it's supposed to be 20% gray. That's why we hold up "18% gray cards" not "50% gray cards" for white balancing. CRTs used to make 18% gray less than 18% gray like 2% gray. So we would jack our 18% grays up to 50% gray so that the defect in the CRT would end up back at 18% gray. 50% gray isn't actually middle gray, it's exactly 1 stop below blown out white. Your graph though starts from linear color space (which it isn't) and goes to less than linear. It should start from the red (gamma corrected) line and map down to the linear line.

/nitpick

No it's good to get this kind of info because I only have rendering engineers to bounce these ideas off of. If I'm going to do more videos I need to make sure I get the info correct.

Thanks

 

[Stacey Spears]

Is the video targeted at computer(s)? I ask because in the computer space it is 0-255 and sRGB. In the CE space it would be 16-235 and BT.601/BT.709. Same for gamma. 2.2 in the computer space and 2.4 in the CE space on the display side.

On Xbox, for example, a game will be remapped from 0-255 into 16-235. For 360 we would twist the gamma. We don't do that by default on Xbox One. 709 and sRGB have the same gamut and white point, so that is not an issue.

Just random notes.

 

[Alexander Mejia]

In the artist centric workflow, 0 is black and 255 is white. If the user is in a TV level space, the chip inside the game console will translate for you. The internal frame buffers are 8-bpc full range sRGB before they hit the scaler chip. What the HDMI Handshake has determined is where it will be translated to.

Because our tools output to three platforms (XBOX, Playstation, and PC) we need a way to work in one color space or else things could get out of control making seperate assets for different consoles. Because of our awesome rendering programmers, we don't have to worry about it.

I wanted to include a section about IRE, but we never see it, so it didn't make it in. A lot of NLE's will take 0 IRE and translate it automatically (kind of like what we do) depending on what you are exporting to under the hood.

 

[Gavin Greenwalt]

Also along those lines, I actually imagine your engine's framebuffer is half float 16bit with 0.0 = black and 1.0 = white. That then is processed in floating point space with your lighting composited in linear gamma and then finally tonemapped to 8bit gamma corrected space and passed to the GPU's scaler. Do any games render in 8bit precision still? I was pretty sure that everyone had moved over to a linear light pipeline, especially with so much deferred lighting going on. It would be next to impossible to do deferred rendering without working in linear floating point space.

 

[Alexander Mejia]

So yes, you are correct in that point where we do math internally at a certain BPC at linear gamma, however when it comes time to the final image we crunch it down to sRGB 8-bpc 0-255 and let the scaler chip take it from there.

You are obviously a little more keen than most out there, but I'm trying to keep the under the hood stuff hidden for those artists who don't need to know.