Colour Management

[Trevor Meier]

Can someone inside RED take us through how colour is mapped from the native sensor colourspace, via RedCine, into available video & data formats and out to the display?

I'm interested in both how to maintain a calibrated preview in RedCine, and ideas on how to maintain a calibrated display throughout post (barring an ultra-expensive solution like TrueLight - though beautiful, mucho $$). Any recommendations for ensuring accurate colour rendition all the way through the pipeline? (monitor profiling, calibration LUTs, getting around Quicktime & FCP's limitations, etc.)


Some side notes:

Currently there's some notoriously frustrating bugs in Quicktime that make gamma rendition inconsistent between inside Final Cut, direct output from the timeline, output via Compressor, and taking files inside of other apps such as After Effects. My particular situation is using Final Cut with two attached Cinema Displays. One example: will doing a calibration of the displays (with one of the various available probes) and using 'Digital Cinema Preview' from inside Final Cut give me accurate colour rendition?

In the canvas window, FCP does gamma compensation based on an assumed 1.8 display gamma and 2.2 output gamma... anyone know if is this true when using 'Digital Cinema Preview'? Any methods for getting around this 'feature'? If anyone has more info on how FCP handles colour mapping on attached displays (i.e. not attached to a video output a la AJA) I'd be very interested to know.

[Graeme Nattress]

Native RED sensor RGB space is transformed to XYZ, and then from XYZ to the selected RGB space, say, REC709. Then, the REC709 gamma curve, say, is precisely added to the linear data, giving a correct REC709 image. Once we pass that image over to the operating system, it's out of our hands though.

We know about the Quicktime gamma nasties and are working to ensure that our codec does not cause issues with this.

Hope that helps,

Graeme

[Corrado Silveri]

One example: will doing a calibration of the displays (with one of the various available probes) and using 'Digital Cinema Preview' from inside Final Cut give me accurate colour rendition?


At this time, no way. FCP via Digital Cinema Preview is NOT an accurate color rendition. The only way is preview your images via Aja/BMD-DVI.

A Graeme article on the subject (using an Apple LCD 23' as a monitoring device via Blackmagic HDLink):

http://www.lafcpug.org/reviews/review_decklink.html

Graeme, is that still valid?

[Graeme Nattress]

For HD, that's correct. The DC Preview is not accurate. I think video has to, for colour calibration, move towards the approach adopted by ICC colour profiles. We can't do that alone, but we do have the science to make our camera work within such a model.

Graeme

[Corrado Silveri]

For HD, that's correct. The DC Preview is not accurate. I think video has to, for colour calibration, move towards the approach adopted by ICC colour profiles. We can't do that alone, but we do have the science to make our camera work within such a model.

Graeme

I'm thinking about this ICC working group:

http://www.color.org/groups.html#digital2

It is a "working" group or not?

[Trevor Meier]

Built-in profiling tools are a big gap in this industry right now... With so many potential colour spaces it's no longer safe to trust the rendition of a particular piece of hardware or software. Digital pre-press went there a long time ago... we need profiling for all things moving digital image.

Slightly OT: Is previewing on a profiled display (using DC preview, or a rendered QT file) going to give an accurate colour rendering for computer-based delivery platforms? (i.e. Quicktime via the web... assuming, and this is a big if, the end-user has a roughly dialed-in display) In other words, does Quicktime obey display ICC profiles & gamma setup (2.2/1.8 etc.)?

OT Side Notes:
* Using adaptive scaling + antialiasing inside compressor can produce some surprisingly good results at 800kbps... though it takes hours to render.
* I've noticed all of the trailers on apple.com are encoded with black at something like 7-10% grey. Is this Apple's way of working around their own bug (gamma issues)?

[Trevor Meier]

Been thinking about this a bit... in digital photography I've got a couple good options as a 'working space' for my shots... Adobe RGB and ProPhoto RGB.

Here's my big wish: Since most of my motion picture material will go out to a variety of final colour spaces (DVD, Digibeta, HDCAM, film, etc), it'd be great if NLE's supported a 'working colour space' that maintained as much colour and dynamic range as is in the image. The NLE would need to obey the system display's ICC profile for converting images from the working space to the colour space of the display. It would need options for 'soft proofing' the video - applying LUTs for sample output spaces: NTSC, sRGB, Rec709, various IP & camera films stocks, etc. On output, it could either convert to (bake in) the colour space, or support any of the upcoming colour metadata standards along with an attached 'working space' profile.

Anyone know if I'm just dreaming?

[GlennChan]

Color management for video is fairly uncomplicated.
A- Most consumer sets have wacky colors, so you can get away with a lot.
Side note: QC reports generally tend not to mention color inaccuracies (i.e. because they tend to stick to objective, asinine stuff), so people aren't getting dinged for inaccurate color.
B- There is only one target gamut. So for monitoring, just get a broadcast monitor and you are done. With HD, this is a lot more difficult as we are moving away from CRT (and CRT's native transfer function) and good HD monitors are very expensive (or not that great quality). But basically this problem is solved with a good HD monitor.

2- There is a movement towards wide gamut / xvYCC for the home theatre/video market, and this will require the monitor to perform color management. IMO, I don't think that xvYCC is a good approach for better image quality and I hope that it will not achieve significant deployment/usage. However, I haven't seen a wide gamut system in practice (other than film projection and slide film, which exceeds video gamuts in some areas).

3- The DI market is already served, albeit many of the solutions don't work with other vendor's products. *Not too sure yet.

4- The DCI / digital projection market doesn't exist yet.

In all the cases above, I don't think working space is really an issue. Your output format has to be a particular colorspace- Rec. 601 / EBU, Rec. 601 / SMPTE C, Rec. 709, DCI (XYZ color space, P3 white point, 2.6 gamma), or a particular film stock. You're only targeting a particular colorspace. (Or did I miss your point?)

[Graeme Nattress]

But 709 is a smaller colour gamut than, say, RED can produce. Therefore by targetting 709 for accurate monitoring, you're throwing away some gamut.

Now, if you kept the image in camera RGB space, and just transformed it to 709 for viewing, you'd not loose anything if you later on wanted to output to XYZ for DCI, or Adobe1998 for print work etc.

We could, like DCI, perhaps, just go from camera RGB to XYZ, (and use a viewing transform to see it as 709 for monitoring, or through to a projector that knows XYZ) but what image processing tools understand XYZ?

Graeme

[Antoine Baumann]

Hello Graeme,

I read your article about the BMD HDLink and Apple 23'' cinema display. I do not want to argue with you, as I do not have a tenth of your knowledge, but you wrote (on that article) that CRT monitor "don't show the full 1920x1080 pixel resolution of HD (due to the physical limits of CRT technology)" and I was sure the Sony BVMA series has 1920x1080 native resolution.
I mention these monitors because some people here in switzerland and france still like to grade on them, and are not trusting the LCD and Plasma monitors. I am not saying they are right, just their experience.

What do you think of product like cine-tal
http://www.cine-tal.com/
or e-cinema?
http://www.ecinemasys.com/products/dcm23/dcm23_features.htm

Thanks,
antoine

[Graeme Nattress]

A CRT just can't be built with the precision to show 1920 pixels across without them blurring into each other. If you measure the resolution of a CRT, they'll fall short of the maximum resolution you need to fully resolve HD. They certainly will accept a 1920x1080 signal, but that doesn't mean that they can fully display all it's detail.

cine-tal and ecinema make great monitors. They're not cheap, but they're nice. I'm hoping that before long there will be cheaper, better alternatives. I think a move to LED backlight on LCD should usher that in.

Graeme

[David Limpus]

Sorry if this is drifting of topic.

Graeme if Canon can now get Surface-conduction Electron-emitter Display (SED) monitors to market. Do you see this as possibly a better colour correction monitor then a LCD led backlight monitor.

http://www.canon.com/technology/canon_tech/explanation/sed.html

[Graeme Nattress]

Well, they could also be good, but we'll have to wait and see and compare when all the next generation of wide gamut displays are available.

Graeme

[Trevor Meier]

In all the cases above, I don't think working space is really an issue. Your output format has to be a particular colorspace- Rec. 601 / EBU, Rec. 601 / SMPTE C, Rec. 709, DCI (XYZ color space, P3 white point, 2.6 gamma), or a particular film stock. You're only targeting a particular colorspace. (Or did I miss your point?)

There's two issues here. One is that there are now several potential target gamuts for every production - sRGB (web), 601, 709, DCI, and film - and no way to know whether what you're looking at is accurate to any of them (unless you pay $$ for Truelight etc.). The second is maintaining maximum gamut from your source through the pipeline, regardless of output colour space. These are both problems that have been solved already with simple, affordable solutions. The colour science & workflow that's already tried & tested for pre-press/still image processing could be extended to 'video' (moving digital images).





We could, like DCI, perhaps, just go from camera RGB to XYZ, (and use a viewing transform to see it as 709 for monitoring, or through to a projector that knows XYZ) but what image processing tools understand XYZ?



XYZ might be a good option, at least for profiling... but you're right, the readily available image tools out there all work in RGB. ICC profiles already use a well-known transfer from RGB space (and other spaces) back and forth to XYZ for applying & transforming between profiles... NLE's could take a similar tack. That would allow data cameras to stay in RGB space, video in 601 or 709, scans in Log space etc. and have the NLE do the conversion. But this requires major NLE support.

In the interim, one option would be to use the same quicktime hooks that allow Final Cut to have the Digital Cinema Preview output on a second display. (It's a quicktime component)

Using some reference images, profile what Final Cut is doing to the image on output to the quicktime framebuffer. Build this into a new quicktime component as the basic transformation LUT, then add a LUT-loader for the output space (a film stock, 709 etc.). Pass that on to the OS which uses your ICC profile to calibrate your output, and you've essentially got a 3D LUT system for any output system. It's two transforms instead of one (so some potential for inaccuracy), but it uses existing technology, and will allow any display attached to a computer to be as accurate as it's capable of.

So all we'd need is somebody to program the Quicktime Component. Graeme, you've got lots of spare time... ;)

[Jeff Brue]

The only way you're going to get any of the features you're talking about are with a professional color management system. Take a look at cinespace, its not nearly as expensive as Truelight and it allows you to switch between Rec 601 and 709, sRGB, and Film stocks. By the way read the DCI spec sometime and you'll see its not really a color space specification as more along the lines of a guideline. Most people are coloring to P3.

[Trevor Meier]

Thanks Jeff... I've got some experience with cinespace, it is a nice product and fairly affordable given the market. With the proliferation of formats and colour spaces, it just seems to make sense that NLE's will incorporate this technology natively, rather than requiring an expensive third-party addon.

[Lucas Wilson]

Thanks Jeff... I've got some experience with cinespace, it is a nice product and fairly affordable given the market. With the proliferation of formats and colour spaces, it just seems to make sense that NLE's will incorporate this technology natively, rather than requiring an expensive third-party addon.

Why would a typical NLE be concerned with critical color calibration? Should NLEs support 1D and/or 3D LUTs? Yes. But in my opinion they could concentrate on getting a true file-based workflow first. Neither Avid nor FCP can support DPX, Cineon, or any real high bit-depth format in anything other than a video space. 16-bit TIFF? 10-bit DPX/Cin? OpenEXR? Nope. And in today's professional NLE market, there are no other real players. Adobe and Sony make great tools (I own a copies of Premiere and Vegas,) but they have zero real market share.

Truelight, RSR, Arri, Kodak, Imagica, and the other companies that do true color management are much more than LUT generators. They provide services that calibrate your viewing environment to a defined set of output parameters, and provide a way of matching your color environment to that of a particular lab for ensuring critical output matching for film out. That isn't typically what editors are worried about, and where NLEs specialize.

...my $.02.

Cheers,

Lucas
-----
ASSIMILATE, Inc.
Los Angeles

[Chris Kenny]

To anyone from the desktop publishing world, the notion of color being separate from editing is very strange. In the DTP world, color management is just a standard feature of every app in the workflow, and, while there are color management experts, pretty much every graphic designer is familiar with the skills and technology behind color management and is making creative judgments about color all the time. In fact, the role of color management experts is not to make creative decisions at all, the way colorists do, but to preserve what the designer has chosen in the final output.

Distinctions between fields which have historically existed primarily because of technological barriers, labor regulations, turf protection within major studios, etc. are already beginning to break down. This sort of thing always starts at the bottom, but it eventually works its way up. An editor is already expected to have the skills of a colorist at the low-end. Don't be surprised if this notion works its way up-market.

And it won't just be about saving money. It's also about art. If a movie is supposed to be a single coherent vision, all the pieces have to come together just right. When the guy editing sequences is also the guy doing grading, that should have (assuming he's actually good at both), a positive impact on the quality of the final product. Presumably the color choices will work with the editing choices in ways that choices made by two separate individuals typically wouldn't.

Based on their purchase of FinalTouch and their longtime focus on color in the desktop publishing world, I'm guessing Apple is going to be out ahead on the integration of grading and editing. Should be interesting to watch.

[Trevor Meier]

Lucas,

I'm making a big assumption in my suggestions that Avid & Apple will move to file-based workflows... to not do so would be against the grain of an industry that's moving to IT-everything.

Assuming we do have file-based workflows, I want to know that my viewing environment can give me a close approximation of that of my viewers... I've had that for years in broadcast, I want it also to be true for web and for theatrical release. If NLE's adopt a data-centric paradigm, the technology exists to give me this for relatively cheap. While it won't be quite a Truelight - calibrated to a print from my lab - it can be close enough, ie calibrated to a generic sample of that stock... or to sRGB... or to 709... etc. each of which I can load as needed.

Even if editors aren't being asked to do more colouring work (they are), accurate monitoring helps in creating & communicating the desired emotional result. E.g. if your gamma is off, the images are too dark or too bright, colours are more saturated than they will be for the viewer etc., it has subtle (even not-so-subtle) effects on how you cut and the shots you use.

[Graeme Nattress]

But 99% of people don't need to go back to film, so no more complexity than what an ICC profile is like, is needed, and that's certainly now doable in real time in an NLE, I'd think. I'd argue that an ICC profile is probably a bit over-complex for the needs of video, but it wouldn't really do any harm.

For long enough, video calibration has been in the monitor, with crude controls. I think that time is now coming to an end....

[Chris Kenny]

I can't really imagine why an ICC profile system wouldn't be good enough even to go to film, assuming the monitor has the appropriate physical characteristics. ICC is used to go from RGB monitors to CMYK printers in the DTP world. That requires figuring out how to reproduce colors using different primaries, and if you've ever seen a lab plot of the color space that can actually be reproduced by a printer, it's usually, well, somewhat bizarre. Going from a monitor to film should be simpler than this.

[Mark L. Pederson]

REALLY COOL new color management tools to be debuted at NAB from Filmlight (makers of Truelight) - and keep your eye on Rising Sun Research ...

[Joe Carney]

REALLY COOL new color management tools to be debuted at NAB from Filmlight (makers of Truelight) - and keep your eye on Rising Sun Research ...

I would like to see, really cool and affordable color management tools to debut at NAB.

[Thomas Mathai]

I can't really imagine why an ICC profile system wouldn't be good enough even to go to film, assuming the monitor has the appropriate physical characteristics. ICC is used to go from RGB monitors to CMYK printers in the DTP world. That requires figuring out how to reproduce colors using different primaries, and if you've ever seen a lab plot of the color space that can actually be reproduced by a printer, it's usually, well, somewhat bizarre. Going from a monitor to film should be simpler than this.

There are more unknown variables when you go back to film, is the "soup" fresh or stale. Are they hitting LAD Aims. It's definitely not WYSIWYG.

[Trevor Meier]

ICC might be overkill (except when calibrating to film stock)... I'd be happy with whatever as long as it's:

a) reasonably accurate
b) affordable

ICC accomplishes both, and the bugs have been worked out...

But, if RSR or Truelight decide to come out with something mere mortals can afford, I'd be thrilled. I doubt it though - in the DI world where Scratch can come in at $30K+ as the value leader (which it is... which is exactly my point...) ... I doubt they will release something at an affordable price point for fear of compromising the huge investment they've made in R&D.

[Blair S. Paulsen]

I think Chris nailed the crux of the biscuit with his analogy to desktop publishing and the concept of barriers being broken from the bottom up. The devotion to established color management strategies might persist in the high end of the market but beyond that forget it. IMHO tape to tape color correction will go the way of the dodo.

Properly set up rooms operated by talented specialists for top quality audio mixing and color grading will still have a place anytime there is a bit of budget and/or quick turnarounds. The real sea change here is that with the maturation of the commodity NLEs (particularly Final Cut) you'll get the ability to get very close to the same finish quality.

We need robust file formats that support higher bit depths and are easily interchangeable. The RedOne > RedCine data centric approach may not be fully supported on day one but given time I expect it to be a very smooth ride into NLEs that are ever more fully featured.

For my part I have painted the walls 18% gray, added some medium gray acoustical panels, spent a little coin on some near field optimized speakers for audio mixing and am ordering an eCinema DCM23 monitor for color grading. I may not be able to match the talent of the best post houses in LA or NY but if Final Cut Studio 6 includes a 10 bit color correction engine...

... and the walls came down, all the way to...

[Trevor Meier]

BTW I'm not trying to devalue what Truelight and RSR do. They've got great science and excellent products. What they offer (specific calibration of environments to lab outputs) is a necessary part of high-end DI workflows.

But I'm not high-end. I just want to be able to see the image in a good approximation of what will be seen by my audience. That isn't the case right now. The technology exists to do it accurately and affordably. So here's hoping someone will do with workflow what Red's doing with camera's and break the artificial barrier that exists at the moment...

[Blair S. Paulsen]

Trevor - agreed, there are some great tools but they are pricey. Right now the desktop / low budget tools for color grading (and profiling devices in the post production chain) for matching to final output characteristics are crude by comparison to what the serious specialty vendors offer.

Having been involved in the DTP revolution I see many parallels here. I think in the near future we will see big improvements in the "desktop" color grading applications. Top talent will still do the best work but self taught colorists, (especially those with Photoshop and/or video engineering backgrounds) with proper monitors in properly set up rooms should be able to turn out beautiful images. I would even go so far as to postulate an expansion in the art form, though along the way we will likely be subjected to some awful dreck. (No pain, no gain.)

Now, how about a decent desktop scope package - the specialty boxes from Tektronix and the like are ridiculously expensive. Final Cut 5 improved their scopes a bit and Scopebox has gotten some good reviews but how about killer scopes in Final Cut 6 - please Steve, pleeeeease...

[Mark L. Pederson]

if Final Cut Studio 6 includes a 10 bit color correction engine...

expect a 32 bit float color engine (on the GPU) but a TON of new bugs that will take 4 or months or so to settle down ...

[Mark L. Pederson]

Now, how about a decent desktop scope package - the specialty boxes from Tektronix and the like are ridiculously expensive. Final Cut 5 improved their scopes a bit and Scopebox has gotten some good reviews but how about killer scopes in Final Cut 6 - please Steve, pleeeeease...

Steve is busy making phones & ipods.

If by killer you mean 100% accurate with amazing detail ... you need to accept that "Killer scopes" are simply not possible without hardware at this time.

OmniTek is the most KILLER. Yes, expensive. But KILLER!

Sometimes you gotta pay to play.

[Chris Kenny]

There are more unknown variables when you go back to film, is the "soup" fresh or stale. Are they hitting LAD Aims. It's definitely not WYSIWYG.

Nor with DTP, where there are variations in ink, paper stock, the performance of specific printers/presses, etc. and ultimately the only proof is, well... the proof. If you're lucky. Yet, the industry manages.

In the film and high-end video production business, in contrast, we have people telling us that color grading is an arcane art which can only be practiced by high priests in tall towers with a quarter million dollars worth of equipment, up to and including a $30K set of knobs to twiddle. Anyone from the DTP world, who has been doing color-accurate work for years with commodity hardware and software, knows that this is complete nonsense. The DTP solutions aren't perfect, but they're good enough.

There are many people who have been in the film or high-end video production industry a long time, and have the luxury of working on big-budget stuff with the best equipment money can buy. Some of these folks seem to have developed the notion that it's not possible to make anything worthwhile with any lesser equipment. You won't see this stated outright often, because when stated outright it's obviously nuts, but it's implicit in a lot of what gets said in some other forums. Not just about color, but about lenses, tripods, audio.... It's something to watch out for.

[GlennChan]

For big-budget work, some of the high-end systems can make a lot of sense. All the other kit costs so much that it makes sense to pay a lot to get a little faster. As well, they also have to keep the clients happy; having to wait on rendering and whatnot is not impressive. And for commercial clients, they sort of want to spend as much as possible to get the best quality. So that's why there is a market for high-end kit.

2- At the lower end of the scale, I certainly think that there is a market for a good desktop color grading program... Final Touch sort of demonstrates this. FT does have its problems though... the main ones being instability, and its poor workflow. Image quality (masking/vignettes/windows look bad) and RT performance also aren't quite up to par IMO.

However, these problems can definitely be solved. Instability, workflow, and image quality can all be solved. And 32-bit FP math can improve image quality somewhat compared to hardware-based systems like a Da Vinci 2K, if you make the leap to linear light processing.

The only shortcoming that may or may not be solved is performance. GPU acceleration might solve this. In existing systems, Final Touch and Scratch (and Lustre) arguably don't have that great performance compared to hardware-based color correctors. Although if you look at Discreet FFI systems (for VFX work), their move to GPU acceleration from SGI shows that GPU acceleration can work very well (although their systems are Linux-based). As well, SGO Mistika (an online system) runs on commodity gaming video cards (it doesn't even need a Quadro workstation card) and runs circles around FCP. But anyways, it should be possible for a really good desktop grading app to exist.

[Chris Kenny]

GPU acceleration will definitely solve the performance issues. The only question is whether it'll do it this year, or two years from now. (I can't imagine it'll take any longer than that.)

And as Motion, Aperture, and a bunch of OS X technologies (Core Image, Core Animation, Quartz 2D Extreme, etc.) show, Apple right on top of GPU acceleration. Final Cut Pro is lagging behind, because it predates the rise of this technology, but it's inevitable that will will join the party eventually. I'd give better than 50% odds we'll see FCP with a seriously overhauled GPU-accellerated engine at NAB, though whether it'll actually be shipping then or only later I won't try to predict.

[Lucas Wilson]

GPU acceleration will definitely solve the performance issues. The only question is whether it'll do it this year, or two years from now. (I can't imagine it'll take any longer than that.)

Not to pat ourselves on the back too much, but SCRATCH can do several layers of realtime grading on 2K images with 32-bit float resolution in realtime... today. Are there things that force a drop below RT? Absolutely... blurs, soft-edge keys, etc. That will only get better, but multi-layers of realtime CC at high resolutions with high bit-depths are here now.


... In the film and high-end video production business, in contrast, we have people telling us that color grading is an arcane art which can only be practiced by high priests in tall towers with a quarter million dollars worth of equipment, up to and including a $30K set of knobs to twiddle. Anyone from the DTP world, who has been doing color-accurate work for years with commodity hardware and software, knows that this is complete nonsense.

Do a grade on a 2048x1556 image in Photoshop.

Now... take 24 of those in a second, and grade all of those.
Now... take several seconds of that shot. It's in the sun. For 5 seconds in the middle of the shot, the sun obscures focus. Make that go away.
Now... there are 15 different pieces of that same scene that were shot on 5 different days with different ambient lighting conditions. Make all of those look like they came from the same 5 minutes.
Now... take those 15 scenes, and make them work in context with the other 200 scenes from the movie.

It doesn't take high priests in tall towers. But is a very, very different discipline. I wouldn't take a colorist and put him/her in the role of an art director at Saatchi, and I wouldn't take the art director and stick them at Company3 grading for Michael Mann.

Maybe it's a lot of the same technical basis, but it's a completely different skill.

[Lucas Wilson]

... But, if RSR or Truelight decide to come out with something mere mortals can afford, I'd be thrilled. ...

Trevor... have you talked to RSR about pricing? They are much more reasonable than you might be thinking.

Truelight - very, very expensive. RSR... very different pricing scheme.

Lucas
-----
ASSIMILATE, Inc.
Los Angeles

[Chris Kenny]

Not to pat ourselves on the back too much, but SCRATCH can do several layers of realtime grading on 2K images with 32-bit float resolution in realtime... today. Are there things that force a drop below RT? Absolutely... blurs, soft-edge keys, etc. That will only get better, but multi-layers of realtime CC at high resolutions with high bit-depths are here now.


Yeah, you guys seem to have a great system. And I'd assume a company with products and pricing as aggressive as yours (and with representatives posting thin this forum) isn't going to be caught unawares by what commodity technology is enabling.



It doesn't take high priests in tall towers. But is a very, very different discipline. I wouldn't take a colorist and put him/her in the role of an art director at Saatchi, and I wouldn't take the art director and stick them at Company3 grading for Michael Mann.

Maybe it's a lot of the same technical basis, but it's a completely different skill.

Sure. I'm not saying it's the same skill. Just that it's a learnable skill, not an arcane art, and also fundamentally something that can be done at an acceptable level with fairly low-cost software and hardware these days (at least, there's no technical barrier), and with even lower cost hardware in a few years.

[Rob Lohman]

Scratch was doing 2K playback in our IBC booth last year. It's a great system!

[Lucas Wilson]

Yeah, you guys seem to have a great system. And I'd assume a company with products and pricing as aggressive as yours (and with representatives posting thin this forum) isn't going to be caught unawares by what commodity technology is enabling.

Nope... we're pretty much constantly aware of it.


Sure. I'm not saying it's the same skill. Just that it's a learnable skill, not an arcane art, and also fundamentally something that can be done at an acceptable level with fairly low-cost software and hardware these days (at least, there's no technical barrier), and with even lower cost hardware in a few years.

Completely agreed. :)

[P Andersson]

I really appreciate the thoughts on color in this thread. Coming from photography, I know the benefits of ICC, it is like the "babelfish" of color. Then TIFF Adobe RGB that acts like the reference base that everybody all over the planet can transform into whatever file format and color profile they want.

Is there at present, a general use file format within digital film/video that irrespective of fps or resolution acts like this TIFF Adobe RGB. It would show up on my monitor looking decent, and would then look similarly decent on another guys monitor if he too has a calibrated monitor, no matter what NLE or operating system was used, and this file containing more color than most other uses, could then be transformed into any other standard for whatever output purpose might be needed.

The thread has talked about that this doesn't really exist yet, so this post is simply trying to establish what would be the best solution "for today" to say move through a couple of NLE's, monitors, operating systems and color graders as smoothly as possible.

[Thomas Mathai]

I see it as that it's easy to get 90% (bad to good) done with any software, the last 10% (good to great) is always the hardest.

I think that's how those people who are used to using the higher end tools see it too. They are used to tools that give them a lot of control, very fast, and they pay for that readily since their deadlines are always fast approaching.

Adobe and Apple have an idea what the high end users want, but at the same time they can impliment only so much at a time.

I think that's why they are making it easier and easier for 3rd parties to fill in the gaps. Any of you checked out Conduit from DVGarage or Colorista from Red Giant Software?

Anything is learnable, it's up to the individual to decide if it's a skill they want to pursue.

A filmmaker I know used Company 3 for color correcting his spec commercial. He said even as they were setting up, with just a few touches, the colorist was already getting closer to the look he wanted. He said he could have done it himself in FCP, but it would have taken him much longer, and he doesn't think it would have been as polished.

[Trevor Meier]

RSR is definitely more reasonable... and it's quite a flexible product (though I'm not sure if it works with NLE's... I'm in contact with them now about that).

Still, though, by the time you purchase a probe and the necessary components, I've paid twice as much to calibrate my display as I have for the NLE suite itself...

Am I asking too much for this to be part of an NLE? Or, at least, for software that's in the hundreds per seat, instead of thousands?

[Thomas Mathai]

Check out Prolost Blog on Stu's thoughts about mastering in your NLE instead of After Effects.

http://prolost.blogspot.com/

[Evin Grant]

What about the new Decklink HDMI Card (Intensity) will this give a color/frame accurate monitoring solution for FCP/AE using LCD computer dysplays ala the BMD Decklink/HDLink option?

[Lucas Wilson]

What about the new Decklink HDMI Card (Intensity) will this give a color/frame accurate monitoring solution for FCP/AE using LCD computer dysplays ala the BMD Decklink/HDLink option?

I guess it depends on what "accurate" means to you. :)

I don't know about the HDMI card... but "accurate monitoring" and "LCD display" don't usually go together. The only two companies I know that are really making serious attempts at creating color accurate LCDs are eCinema Systems (www.ecinemasystems.com) and Cine-tal (www.cine-tal.com).

Lucas
-----
ASSIMILATE, Inc.
Los Angeles

[Trevor Meier]

Check out Prolost Blog on Stu's thoughts about mastering in your NLE instead of After Effects.

http://prolost.blogspot.com/

32-bit rendering is definitely a good reason to go to AE for mastering. But that still doesn't negate the need for accurate colour in AE (and Final Cut, PPro, Combustion, etc.) all the way through the pipeline.

I've had long discussions with a friend from a local DI facility... for what I'm asking for (a 90% solution) a 1D LUT system like RSR's EqualEyes looks like the best bet... I'll quit ranting on this subject since a good solution does exist... but I'll sit quietly in my corner and mumble about how many few-thousand-dollar purchase I have to buy to get a complete workflow.

I suppose it wouldn't be as much of a complaint if I didn't keep agreeing to do these great projects I'm working on that are mostly for the love and little for the money. My choice!

[Blair S. Paulsen]

Can anybody post a link to the RSR color management product. My googling has yielded nada.

[Hrvoje Simic]

here you are

http://cinespace.risingsunresearch.com/product.html



this might be interesting regarding panels

http://catalog2.panasonic.com/webapp/wcs/stores/servlet/ModelDetail?displayTab=O&storeId=11201&catalogId=13051&itemId=102542&catGroupId=14624&surfModel=TH-50PF9UK

they now state 4096 steps.

I've seen the consumer version which got the EISA award and the colors are amazing. No lcd I've seen can top that.

I'm really interested to check out OLED's when they come out.

[Joe Carney]

Hope I haven't posted this before, but here is one from JVC that also displays area markers for different formats.

http://pro.jvc.com/prof/attributes/features.jsp?model_id=MDL101631&feature_id=01&itempath=null

msrp 4695 with both dvi and sdi inputs. Native 1920x1080 and 1280x720p with one to one pixel ratio.

Joe C.

[Jeff Brue]

As a person who comes from a bit more of the indie world, and is in the middle of learning scratch. I do have to say that the art form of real time compositing and color grading is just that, an art form. There are some people who can learn it and some people can't. When you get into the arena of dealing with clients and or having them ask you to pull a key on a girls body color , darken it up, and then take out the grain on her skin, but leave it in for the background. Then hit play and have it render in real time while they watch... lol you're playing several different roles artist, client manager, and IT engineer. Thats why a good colorist is paid comensuratly.

In regards to color management its the same game, film outs, DCI packages, to youtube video releases. They're all different, and yes you will have a client whether you're shooting wedding videos, or a spot for nike ask you why their dvd looks so different from their web copy.

Right now RsR is what we're looking at just because of the price point/ level of service. It'll at least get you in the ball park and in the game.

[Petros Nousias]

Lets say you have an eizo colouredge monitor(i believe thats their top of the line monitor series) and plug it to an aja using an hd-sdi to dvi converter. would you have a colour accurate monitor or not? if this monitor is as good as its supposed to be, whats the difference colourwise with one from ecinema and the likes?

[Trevor Meier]

An Eizo (a cinema display, or other high-end LCD) will give you reasonably accurate colours... but the question is, accurate to what? NTSC is different than Rec 709 (HD) is different than film, etc. Each is unique in how they store and respond to colour.

The solution you mention - AJA HD-SDI to DVI converter - will likely have settings for Log footage, 709, etc. It is a reasonable way to go, but it has limitations in usefulness. E.G. how do you calibrate for your particular display? Also, most of these devices use 1D LUTs which will get you close, but can't accurately represent mediums that use a different display technology (e.g. LCD vs. film).

[Petros Nousias]

I know that its a bit off-topic, but i have a certain issue with an upcoming project to be shot with the hvx-200. We will be editing and colour-correcting in a macpro propably with final cut for both. We can afford an eizo colouredge monitor for viewing purposes through the second dvi output of the nvidia card(dont remember which). A friend has a gretag-macbeth probe so we'll calibrate it as well. so the question is, since this will be for tv only(satelite hd) is there any need to get a multibridge card(again connect the monitor through the card's dvi output) to be colour accurate, or not?

[Manfred Lopez]

Does anyone know how much a Truelight system costs?

[Trevor Meier]

Does anyone know how much a Truelight system costs?

The old adage... "If you have to ask..."


since this will be for tv only(satelite hd) is there any need to get a multibridge card(again connect the monitor through the card's dvi output) to be colour accurate, or not?

If you're choosing to do colour correction on your NLE, you'll need some way of mapping what that display can do, to rec709 (HD) colour space. Rising Sun Research has Equal Eyes, which is a software-only solution. Other 1D-LUT options are the multibridge or the Matrox MXO. My suggestion would be an Apple Cinema Display plus using one of these solutions, which should run you about the same as the Eizo monitor.

[david farland]

Can anyone tell me the difference between 1D,2D & 3D luts? Also digital meaning 1 & 2 light passes? Are they related?
DF

[GlennChan]

LUT = look-up table.
http://en.wikipedia.org/wiki/Lookup_table

(From what I understand, correct me if I'm wrong.)
For a 1-D LUT, you take the input number, find that number in the table, and use the associated number. So you might have something like...
00--> 00
01--> 02
02--> 04
03--> 06
04--> 08
05--> 10
(etc.)

A LUT like that would make the whole image brighter. You can apply a 1-D LUT on the red, green, and blue channels. You can do things like adjusting the white point/balance of the monitor, change the gamma, etc.

A 2-D LUT is like a 1-D LUT, except in two dimensions. You might visualize it like a chess board, with 8 x 8 entries. You would number/index the board along its width and height. Each square on the chess board would contain a output value.

A 3-D LUT occurs in three dimensions. You can visualize it as a cube. Each cube within the whole cube would contain an output value.

2- The 3-D LUT is the most sophisticated and powerful of all. It can describe transformations that 1-D and 2-D LUTs can't (i.e. color response of film).

3- In practical implementations, 3-D LUTs run into a problem. If is a LUT with 8-bit entries, then its size is:
256 x 256 x 256 x 2 = 33,554,432 (bytes)

That many entries won't fit into the CPU's cache, so it would have to fit into the RAM. The problem with that is that accessing data from the RAM can be very slow- it can take several hundred CPU cycles. So to get around this, people use less bits in the LUT. This reduces the accuracy/precision of the LUT. There are some ways of squeezing a little more accuracy/precision out of a particular 3-D LUT. One of them is trilinear interpolation.

[P Andersson]

3- In practical implementations, 3-D LUTs run into a problem. If is a LUT with 8-bit entries, then its size is:
256 x 256 x 256 x 2 = 33,554,432 (bytes)

That many entries won't fit into the CPU's cache, so it would have to fit into the RAM. The problem with that is that accessing data from the RAM can be very slow- it can take several hundred CPU cycles. So to get around this, people use less bits in the LUT. This reduces the accuracy/precision of the LUT. There are some ways of squeezing a little more accuracy/precision out of a particular 3-D LUT. One of them is trilinear interpolation.

i don't know if i remember right, or was there a reason that the DCI specs didn't want to use ICC profiles for this reason, that it would take too much time in processing power before the image gets sent to the projector

[Chris Kenny]

i don't know if i remember right, or was there a reason that the DCI specs didn't want to use ICC profiles for this reason, that it would take too much time in processing power before the image gets sent to the projector

Hmm... this could be solved easily by creating a copy of the media with the profile already applied, instead of trying to apply it to a "fresh" copy in real-time. Perhaps they didn't want to introduce that sort of complexity.

[GlennChan]

I don't believe ICC profiles are necessary since the target is always DCI's X'Y'Z' space. Since there is only one target (as opposed to multiple possible targets/inputs), ICC profiles aren't necessary.

It is up to the playback system to map from X'Y'Z' space to the display device's color gamut. It can use whatever system it wants for that.

[Chris Kenny]

I don't believe ICC profiles are necessary since the target is always DCI's X'Y'Z' space. Since there is only one target (as opposed to multiple possible targets/inputs), ICC profiles aren't necessary.

It is up to the playback system to map from X'Y'Z' space to the display device's color gamut. It can use whatever system it wants for that.

Well, sure, you don't need to embed an ICC profile in a DCI movie, because you know what color space it's in. ICC profiles would be useful for that second problem, though, the mapping of that X'Y'Z' space to the display device. It would be been nice to have a standard approach to this based on a widely deployed technology, instead of leaving the whole thing up to device makers.

[Petros Nousias]

The reason I was going for the eizo instead of the cinemadisplay is because ive read that its far more accurate. Any thoughts on that?

[Graeme Nattress]

X'Y'Z' was chosen to avoid the need for metadata to describe the source RGB space, rather than for computational efficiency.

If you have a reference R'G'B' with known 3x3 transform to XYZ and a white point, it would be:

R'G'B' -> (lut) RGB -> XYZ -> Transform white point if needed -> Target RGB -> (lut) target R'G'B'

with DCI it's

X'Y'Z' -> (lut) XYZ -> Transform white point -> Target RGB -> (lut) target R'G'B'

Graeme

[GlennChan]

1- I think the problem with R'G'B' is not because you need the metadata, but because R'G'B' isn't wide gamut.

2-
ICC profiles would be useful for that second problem, though, the mapping of that X'Y'Z' space to the display device. It would be been nice to have a standard approach to this based on a widely deployed technology, instead of leaving the whole thing up to device makers.
I think the problem with ICC profiles is that the LUTs are all in LAB space, and that LAB is too computationally expensive for video work.

2- The other aspect of DCI is that you don't really need a standard on the color management end. If you implement some proprietary color management system, that won't affect interchange with other users.

So suppose you have a projector with six primaries (I think TI has something like this). The additional primaries help extend your color gamut.

You could implement your own color management system, where you have 3 primaries in (DCI/X'Y'Z') and 6 primaries out (the projector's primaries).

[Graeme Nattress]

R'G'B' can be as wide a gamut as you want - just depends what you pick as your primaries. XYZ gamut is as wide as it gets, but can't be monitored. Camera gamut is again, less than XYZ.

Yes, LAB is computationally intensive. ICC PCS is XYZ D50, and ICC does spec RGB -> PCS (XYZ) -> RGB without going near LAB though.

You can allow the projector to convert to it's R'G'B' from XYZ via it's own method whether you give it R'G'B (in a known space) or X'Y'Z' with white point E. The difference is really metadata.

I see the DCI "point" about metadata, but seeings as the rest of the format is totally reliant on metadata, it's mad that they use the "no need for metadata" as a reason for the colour space for encoding.

Graeme

[Mark L. Pederson]

Lots of anti-DCI feelings building in the industry - especially overseas -

The whole "academy screenings" only in DCI equipped theatres - was a giant disaster in NYC -

I don't thing the jury is out yet on the "real future" of digital cinema in terms of exhibition -

[Graeme Nattress]

For one, it's got to be cheaper, easier, and designed so that "everyone" can be part of it. DCI specs, to me, are a compromise, and that they're really designed for film scanned to a DI, rather than digitally originated material.

[P Andersson]

zodiac cinematographer savides in an interesting article

Savides was asked to create a series of templates such as Day Exterior, Night Exterior, Day Interior, Night Interior — which he refused to do. He explained, “I couldn’t look at them. I didn’t want the look-up tables to bias my eye. I wanted to work with a neutral slate, and that neutral slate had to be that RAW file. It’s the only way I could understand what I was doing everyday. The look-up table would slant you toward whatever you made that look like"

or

“He’s amazing. I don’t think anybody could’ve done it this way. David had to figure it out on his own, and then present it to the studio. He had to do smaller projects, commercials. He’d been using the Viper, got really used to it. So by the time I stepped in he had gotten the Viper integrated and he’d figured out how to make the camera work. When I got there, 90 percent of the problems had been ironed out. I was just part of the creative solution.”

http://www.filmmakermagazine.com/winter2007/line_items/there_yet.php

[David Kirlew]

RSR is definitely more reasonable... and it's quite a flexible product (though I'm not sure if it works with NLE's... I'm in contact with them now about that).

Still, though, by the time you purchase a probe and the necessary components, I've paid twice as much to calibrate my display as I have for the NLE suite itself...

Am I asking too much for this to be part of an NLE? Or, at least, for software that's in the hundreds per seat, instead of thousands?

Trevor, check out Color Symmetry at www.colorsymmetry.com They are reasonably priced, and I've been in talks with them on a project I'm in talks on. Color Symmetry has direct support with 3D apps and I've been talking with them about support for different NLE's (the three A's).

[mezmo]

I think Chris nailed the crux of the biscuit with his analogy to desktop publishing and the concept of barriers being broken from the bottom up. The devotion to established color management strategies might persist in the high end of the market but beyond that forget it. IMHO tape to tape color correction will go the way of the dodo.

Properly set up rooms operated by talented specialists for top quality audio mixing and color grading will still have a place anytime there is a bit of budget and/or quick turnarounds. The real sea change here is that with the maturation of the commodity NLEs (particularly Final Cut) you'll get the ability to get very close to the same finish quality.

We need robust file formats that support higher bit depths and are easily interchangeable. The RedOne > RedCine data centric approach may not be fully supported on day one but given time I expect it to be a very smooth ride into NLEs that are ever more fully featured.

For my part I have painted the walls 18% gray, added some medium gray acoustical panels, spent a little coin on some near field optimized speakers for audio mixing and am ordering an eCinema DCM23 monitor for color grading. I may not be able to match the talent of the best post houses in LA or NY but if Final Cut Studio 6 includes a 10 bit color correction engine...

... and the walls came down, all the way to...

I don't know Blair,
Getting an editor to colorgrade is a bit like getting a vet to do your
dental work.
much pain involved.
Mezmo

[Joe Carney]

The reason I was going for the eizo instead of the cinemadisplay is because ive read that its far more accurate. Any thoughts on that?

Petros, I'm looking at them too. They have hardware based color correction (on board monitor..) 12bit lut and 16bit processing.
I've been looking at this one...
http://www.eizo.com/products/graphics/cg241w/index.asp

and wonder if it couldn't be programmed to give the 'look' you want.
I wouldn't use an Aja either, just a quadro fx or FireGL. Seems for those on a budget, a reasonable solution for playback/preview. (2 monitor system).
My goals are Broadcast, Blu Ray and Internet. When I shoot 4K it will be downrezzed to 2K or 1080p for everything.

[M Most]

My goals are Broadcast, Blu Ray and Internet. When I shoot 4K it will be downrezzed to 2K or 1080p for everything.

In that case, you'd be much better off with a broadcast type monitor (as opposed to a computer monitor), as that is already displaying Rec709, which would be your base color space for both of these applications, and thus does not require the types of calibration you're talking about. Of course, you will need to feed it HD video, so you'll need a card for that (either a Kona or a Blackmagic will do).