How To Author Calibration Image For Video Game

Honestly I find hard to believe that a studio working on PS3/360 games would be this clueless about calibration... or are those simple PSN/XBOX Live games?

Well the most important thing is that every single screen being used for graphics development and testing needs to be properly calibrated: http://www.avsforum.com/avs-vb/showthread.php?t=948496

Then you could provide basic black level pattern in game. A gamma patern is pointless since its sweetspot is greatly affected by room lighting, etc. so you could simply provide a gamma slider and tell the user to adjust it to its liking after adjusting black level.

Besides providing those two very basic patterns I mention you really shouldn't bother much about user settings, once your development/test workstation screens are properly calibrated to the standard it's up to the end users to do the same on our side to replicate what you guys intended to show us.

Then again, maybe simply mentioning that users who want to make the most out of the visuals should try using AVSHD 709 on the game manual would be the ultimate gift for people who don't know about screen calibration.


So can you give us a hint on what company you work for?

Okay sounds like you need a quick primer.

PC levels 0-255
TV Levels 16-235

I'm not sure what goes on with xbox or PS3, but black when output to a TV should be at 16. I don't know the guts of those boxes well enough to know if the console compresses the range on output or if developers need to author their content for 16-235.

I do know I've seen some games on the PC where the full motion video was authored for 16-235 and thus black level was too bright on the PC.

I'm no game developer but I read a document authored by valve software that discussed the xbox 360's non linear gamma curve and how to get the best results from it, specifically with regard to keeping your games from looking to dark. I think it may be worth at least a look.

http://www.valvesoftware.com/publica...eOrangeBox.pdf

The console system itself probably takes care of the conversion from 0-255 to 16-235.

Sorry if I sounded a bit on the harsh side. My point really was based on the fact that game developers so far don't seem to give a damn about picture standards, or at least won't let us know if they do.

An example of this is how the PS2 (at least the pre-slim ones) output levels are fixed at 1-234 over component, no option to change it. These days every TV can take PC levels, but back then in the CRT era you had to take the black crush.

It would be just awesome to see proper patterns included on a PS3 title, I've always wanted to know if a video based calibration holds up on game mode. If you guys manage to include some sort of calibration wizard that passes the mom test on a AAA console title you'll make the video world better for everyone.

Wolfenstein huh? Sweet, I can see why you guys care about blacks with all that cool green undead glowing thingie you are using


This article is a great read: http://www.curtpalme.com/forum/viewtopic.php?t=10457

You need patterns for pc levels and tv levels. Nothing special should be needed from the artist. However, i would be careful how much control you give the user, not all easily understand even patterns, and it could possibly make things worse.
And, Edo, it is actually not that uncommon. I do pc games as an artist (i do contract stuff, but no professional studio work), and calibration/brightness levels are really no concern to anyone, except maybe the scene construction artist, who wants their lighting to be consistent.
Although, one thing i've always wondered with the pro studios, do they calibrate their own monitors? If not, the lighting will be inconsistent anyways, and it won't matter what controls you give the user, the end result is unlikely to match the original artist's intent and vision.

The pluge pattern the use for TV calibration is generally something like this

below black bar at 0
background at 16 -- 0%
above black bar at 20 -- 2%
above black bard at 25 -- 4%

Then you tell the user
0 bar = invisible
20 bar = barely visible
25 bar = clearly visible

Actually they don't have to worry about pc and video levels, the PS3 will map game content to whatever you have it configured for, analog/digital component, full/limited RGB, RGB(0 0 0) will be either a 1 or a 16 depending on the output format.

To complicate matter even more, both the PS3 and 360 allow users to select either PC (0-255) or Video (16-235) levels. So you have to account for both scenarios. On the 360 there is even a 3rd choice called intermediate which uses who knows what levels (something in between PC and Video levels).


The answer to this will be to determine what PC level your rendering output uses as "black" on each console. A calibration screen is useless if you don't have the proper reference to calibrate to. So you'd have to figure out what level "black" is rendered to on each console and then base your calibration pattern around that. If your rendering output ends up using something above PC level 0 as black, you can potentially use some form of modified pluge pattern. If you end up rendering with black being PC level 0, then you will have to have a "make sure you can distinguish this box from the darker box inside it/next to it/etc." type pattern with PC level 0 being the darker box.

If you can, you might want to try and contact another developer that has already dealt with this situation and see what they have to say about it. The fact that the consoles will eventually output at either PC or Video levels no matter what levels you render to has to have added some level of complexity to this as well.

Yes, but this conversion is handled by the console. The developers don't have to worry about this. After all, the 360 didn't even allow 0-255 output until more than a year into its life.

Can I get my $10000 color pipeline consultancy fee before I give you any advice?

Game devs ... are rubbish about video standardisation. I work with lots of artists who migrate from games to films and I often go through following scenario.

CGI arrives:

Me: what colorspace is this?

Artist: its linear.

Me: what do you mean by linear?

Artist: its linear video.

Me: is that linear or video gamma?

Artist : its gamma 1

Me : you just said its video

Artist: it is video and its gamma 1.

Me: what's this supposed to look like?

Artist: whatever looks best to you.

Me: do it again and when you can tell me what colorspace it is I'll actually bother to look at it.

This is such an easy topic that gets ridiculously over mystified.

3d lighting is a linear model: just like the real world. The maths is all linear.

What makes the end imagery a given colorspace is the rendering intent.

This means whatever the lighter ( and texture artist) was looking at when they created the imagery. .

For example if they used a nominal sRGB monitor with a end display gamma of 2.2 and no additional display correction or color management anywhere in the pipeline the resultant imagery will look subjectively correct when viewed back in the same display enviroment. Even though the artist has just been rendering with linear maths what they are actually creating is imagery that has a gamma of the inverse of 2.2: ie 0.4. This is kinda like a video gamma (I'm simplifying here , the point is though that the resulting image gamma is not linear : ie 1)

If you wanted an artist to create imagery that atually had a gamma of 1 they would need to do the lighting with an end display enviroment that was itself linear (1).

How you give them this sort of enviroment can vary:

The most common way is to apply the inverse of the display hardware's gamma to the graphics card . Then you have a graphics card doing 0.4 and a display doing 2.2 and which gives you a working enviroment of 1. (this is an oversimplification)

You could if you wanted then render your linear CGI through a LUT ( in a shader for example) that transformed your linear cgi into video. If you do this perfectly you could then display your imagery in a video enviroment and it would look exactly the same as it did on your linear environment.

The way to conceptionalise this way of working is that the CGI is actually trying to recreate a real world linear scene ( like real life) and your final shader LUT is actually an analogue to the way a video camera (or film or whatever notional capture system) would make that scene look like if it was captured by said device in real life.

This is where things like HDR become useful because you are rendering a large ( notionally infinite) dynamic range and then making decisions to capture or map some of that range into a smaller exposure envelope mimicing the transfer characteristics of a real world mechanical system: ie a video or film camera. But obviously because you have to carry out the lighting calculations in HDR you have significant overhead compared with just rendering what actually ends up being visible ( of course there are all sorts of cheap effective ways way to limit the actual data requirements).

So in the case of rendering for a "video" console:

What colorspace is the end output?

Lets assume for an video console with 8bit output thats a display with a gamma around 2.2 and a black ref of 16 and a white ref of 235.
So your end imagery needs to be 0-255 designed to look correct if everything below 16 is clipped to black on display and everything over 235 is clipped and has a notional gamma of the inverse of 2.2 ...ie 0.4.

However is you want to differentiate your imagery from not looking rubbish like your competitors ensure that you render intensity variation all the way from 1-255. I actually keep everything in the whites , if you think about how white ref scales against the real world peak white point of a given display its easy to see why.

How you get from a linear lighting model to 0.4 is really down to you : apply an end LUT to take 1 to 0.4 or just light the thing whilst looking at a video display enviroment with no correction in the pipe.

How the 360 behaves is something I'm sure is easy enough to find out maybe it assumes all 3d is linear then it just LUTS it on output (doubt it) ,maybe it doesn't touch it, maybe it expects a PC 0-255 range with a 0.4 gamma that it remaps to video , maybe it does this dependant on the actual hardware output type. ( It would actually be a lot more straight forward if the 360 is totally transparent and then you guys would know exactly what your target was...I'm not saying this isn't the case)

You also need to realise that "gamma" itself is a simplification and to be really accurate ( and probably make your imagery look better than your competitors) you want to deal with more accurate curves in your LUTs that properly reflect real video colorspaces. Especially behavior towards blackpoint and whitepoint.

Some devs I thnk get it right:

Bungie ; all the Halo games look as if they are outputting video levels to me on a 360 with a component connection to a properly calibrated video display.

EA: COD4 looks right on the money to me. So does Mass Effect

Some devs I think get it wrong:

Ubisoft: Far Cry Instincts looks like PC levels so on a video dispay the blacks look crushed.

King Kong had a well publicised issue with looking too dark that ubisoft (rather pathetically) blamed on the 360 hardware rather than their own lack of defined pipeline.

Bethesda: Oblivion looks like PC level to me: Fallout 3 looks like video ...maybe they realised something.

If all else fails you can stick a gamma or brightness slider on your games: a massive telltale that a given developer has no confidence in their color pipeline if ever there was one. Although its possibly not a bad idea assuming you actually get the deafult levels right in the first place and give people a meaningful pluge to set it by.

When's the ;last time you saw a dvd with its own gamma and brightness settings

Well you won't have an accurate reliable useful competative advantageous image pipeline and your imagery will not look as good as your competitors.

Its like trying to compete on the Formula One circuit and putting remoulds on your car.

Hi Manveer, having spent the better time of my working career as a videogame journalist, I was discussing with a colleague about your request.

We tend to agree with Edo Galvez on this one but would like you to consider an extra step. Once you have had the players calibrate their screens with the pluge pattern you were inquiring about, it would be great to have the following.

In the pause menu there should be a gamma setting to adjust gamma properly for dark scenes rendering (as mentioned, this varies according to lightning conditions in the players' room). This gamma setting should be adjustable with the paused picture of the game in the background without any of the usual dimming that happens when you pause the game. Once you select the gamma adjustment from the menu, you should have a small gamma slider in view (preferably with some kind of transparency) and the whole game picture to check against your adjustments.

This way, after having calibrated black and whites (you never mention contrast but it's important as well not to have white crush), people would be able to adjust interactively with real game content.

I hope this helps. Cheers and happy developing.

Like I said a company with confidence in its image pipeline should have no need to add additional sliders to their games. Gamma tweaking based on lighting differential in the viewing environment should be handled at the display not the material.

The color standards employed by the average game dev are frankly amateurish. Sticking pluge patterns and sliders onto games just makes me shake my head every time I see it. However if you must do it at least make it meaningful.

When you put sliders on your games basically you are saying to the customer,

"we don't know what we are doing , we don't care what our imagery looks like and we are passing the buck to you"

The most agravating thing about this is that its EASY to get right. The concepts are simple, the maths is easy, the day to day maintenance is quick and you can buy inexpensive software off the shelf to do this with great support.

Let's assume a couple of things.

The Xbox 360 doesn't do any remapping and is essentially transparent to what you tell it to render.
The Xbox360 is a video games console not a PC so lets assume its going to be hooked up to a video calibrated display in actual use.

You have a bunch of lighting and texture artists appraising CGI on a workstation . You need to ensure they are all doing final appraisal for the same target for high def video , that's Rec.709 (nominally gamma 2.2 on display , black and white ref is 16-235 for 8bit.

CGI lighting calculations are a linear model. This makes no assumption as to the rendering intent (rec.709).

Step1. Consistent display enviroment.

Every combination of display and workstation will have different image characteristics so you need to make your baseline consistent. This involves running a closed loop calibration to profile/measure each workstation . This allows you to then apply correctional LUTS to bring all the workstations into the same baseline. This would allow you to then take each display to the same standard and also they would then referentially alll look the same ( you don't want artists moving from one machine to another and seeing different imagery).

Step2. Where to impliment the video transform.

This is really dependant on how your rendering pipeline works.

Simplest method.

Now all your artists are on calibrated workstations giving them a baselne rec.709 environment . Your renderer is working without any notional connection to the end rendering intent and what makes it look correct for your end rec.709 target is your artist eyeballing your imagery and making decisions as to the final look on a rec.709 workstation.

End result: rendered imagery looks correct on a Rec.709 display.

More complex method:

Instead of taking your workstations to a rec.709 baseline you take them to an intermediate linearish one with suitable D.65 primaries. To twist your CGI into your final rendering intent you apply a LUT (or color tansform) as a shader to take linear imagery into Rec.709.

End result: rendered imagery looks correct on a Rec.709 display.

Considerations:

How accurate do you want to be? Just gamma and black and white ref or do you want to ensure your white balance (grayscale) and gamut ( color) is also accurate. I suspect most games companies only worry about the first two , the better ones worry about white balance and hardly any worry about gamut.

Hardware LUTS applied to graphics cards are usually only 1d models which won't properly correct gamut but 1d hardware LUTS are easy to impliment.
You could use the hardware LUT to get to a baseline and then apply a secondary shader to provide further "3d colorspace" correction to make just the gamut accurate.

Most 3d packages don't allow for the internal application of software display LUTS (more accurate as you can apply 3d color models that will also correct gamut). This actually makes sense as the final imagery should be notionally correct without any additional correction beyond the actual display characteristics ie video, rec.709.

This can be a good way of working as you may also otherwise hit limitations on the actual display hardware: you may start to impact precision if aggressive hardware correction is required for a particularly poor display ( solution...buy a better dispay).

Also your graphics engine can do all sorts of funky things if you impliment final color correction with shaders . All those HDR realtime exposure shifts , B/W to color transitions , Hallucination effects. (Most 3d game engines make extensive use of shader these days). You can make final look changes with a simple shader rather than have completely change the lighting. You can have specific shaders for artists working on specific games scenes that maybe have different creative color considerations without having to change your baseline company calibration.

It can also be useful for artists to have a selection of luts in terms of brightness stops to allow then to see deeper into the darks or whites especially if they are working HDR with an engine that's going to dynamically change the delivered exposure range according to game action ( gunfire flashes lighting up caves for example).

Also you may want to consider just the hardware correction route for most artists for actual interactive work and then a more accurate workstation for final appraisal.

So not only is it easy to get right its also advantageous from a practical and creative perspective.

There are various 3rd party color management systems that will provide this functionality. The larger companies probably build their own to a certain extent. I've used a lot but curently I like Cinespace from Rising Sun but I'm generating imagery for feature films which is arguably a bit more complex (same principle though).

Again if the 360 is doing some sort of remapping its something you can impliment it in a shader. If the architecture is smart you could even swap in a suitable corrective shader to handle whether the 360 is doing video or PC levels on output.

Every company does things a little differently but those are the general principles.