Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Final Cut Pro X Released

Tim Morten said:
A year at the top on the strength of a best-selling phone isn't a reason to deify Apple. They make mistakes like everyone else, and just like Microsoft, they won't stay at the top forever. Android is already outselling them in the phone space, and the tablet market is WAY too young to call. Apple remains at less than 10% of the PC market. iPod is their one uncontested win.

The Apple adoration runs thick with some, bordering on religious; Apple does some good work, and some not so good work. This thread is about a product that many of us consider to be in the latter category.

Personally, I think Apple is a better company when they are the underdog. Let's see what the future holds...
Click to expand...

" A year at the top..." ??? Seriously? Did you really say that? LOL
 
I don't think there's any doubt that Apple is by far the biggest tech company in the world. They are completely dominating the tablet market with a 80% share for the iPad. And while Macs have a 6% share or so worldwide, you also have to consider that a) they are up against a hundred PC manufacturers + custom builds and b) they don't sell cheap computers. Considering these, 6% is a pretty incredible market share.

Commercially, Apple is right at the top, I don't think there's any debating that.
 
Andrae Palmer said:
How would color calibration work with Colorsync. I'm genuinely curious. Say you get a colorimeter and it does a color profile of your display... you then load that profile for your display... how do you then do the rec709 conversion?
Click to expand...

This is what ColorSync does. Given a profile of a physical device (provided with the device or, ideally, generated through measuring that specific unit's performance), and an image in a standardized color space like Rec. 709, it just takes some clever math to display that image accurately, assuming the physical device has a color space as large as or larger than the standardized color space.

Andrae Palmer said:
I'm curious about this possible ColorSync solution. I think like most people that the graphics card will muddle the signal. I know you discussed this elsewhere Chris (stumbled on your thread over at creative cow). I would like to prove once and for all if this srgb to rec709 conversion for color accuracy is possible. Not to prove you wrong... but to learn what the issues are. So far in my research I've read that srgb has slightly varied gamma compared to the gamma of rec709. I've read that the processing on the graphics card will make the signal inaccurate. You seem like a logical person by reading your posts on here. I would like to know how you counter these arguments.
Click to expand...

ColorSync does things like trying to emulate the color spaces of CMYK printers (which tend to be crazy and non-linear) on RGB screens. Displaying Rec. 709 on a computer monitor is nearly trivial in comparison. Essentially, ColorSync is a GPU-accellerated 3D LUT engine, living in the operating system.

There is nothing inherently inaccurate about color output through GPUs. If I recall correctly, there was an implication made in that Creative Cow thread that desktop GPUs punched up images in arbitrary ways to make them look better or something. This is absolutely not true. The reason color accuracy on screens hooked up to GPUs has traditionally been worthless for video is simply because there was no color-managed pipeline for video. Even if you hooked up a screen like the Dreamcolor (which can internally calibrate itself for Rec. 709), you had no idea what you were getting, because the editing software and the OS didn't just pass raw RGB values through to the screen. FCP 7 tried to do a quick and dirty compensation for video/display gamma differences (based on assumptions about your display, not its ColorSync profile), and then ColorSync, treating what FCP 7 was displaying as image data with no color profile, simply applied whatever standard correction was implied by the display's profile. This produced a result that didn't accurately represent the RGB values in the original image (unlike output via a video I/O card) and had no particular relationship to an accurate video image in terms of its appearance on-screen (unlike actual color-managed output), the worst possible sort of signal you could have for color-accurate monitoring.

With ColorSync support in FCP X, the system now has a profile of the display, knows what color space the source material is supposed to be in, and can map between these. It's not different in principle from calibrating a broadcast monitor using an external LUT box, except it substitutes software and mass-market hardware for several thousand dollars worth of specialty gear.

More generally... there seem to be a lot of people who think there's some special magic going on with broadcast monitors. There's not. I guarantee that until you get into very high-end gear, broadcast monitors are using the same sorts of panels as computer displays. The only difference is the on-board image processing... and any digital image processing that can be done by hardware can be done by software. There's no magic here. It's all just manipulating ones and zeros.

Of course OS X doesn't support 10-bit output through GPUs yet, which is annoying (as a bunch of GPUs and a few monitors now support 10-bit). But I've done A/B comparisons of 8-bit vs. 10-bit output through Resolve... the difference is invisible on most images, and even where it shows up (as banding in smooth gradients in the sky, for instance), I can't imagine it influencing actual grading decisions, really.

Mind you, I haven't yet done side-by-side tests between FCP X's display on a calibrated desktop monitor, and the image we get through our more traditional monitoring pipeline. I can't vouch for Apple's specific implementation. But there is no fundamental technical barrier to accurate display of Rec. 709 images on modern high-end desktop displays.
 
Chris Kenny said:
Of course OS X doesn't support 10-bit output through GPUs yet, which is annoying (as a bunch of GPUs and a few monitors now support 10-bit). But I've done A/B comparisons of 8-bit vs. 10-bit output through Resolve... the difference is invisible on most images, and even where it shows up (as banding in smooth gradients in the sky, for instance), I can't imagine it influencing actual grading decisions, really.
Click to expand...

Well, perhaps you can't imagine it, but I've been in sessions where the differences were real and problematic for the grade. If a blue sky is banding, do you let it just band or do you manipulate the grade to avoid that? Perhaps more importantly, since darker colors are starved for bit depth in 8-bit color, how much do you care about banding in the shadows? And if you can't see the real result, how do you make decisions about how far you can push a grade? Chris, I don't believe for a minute that you believe this statement.
 
Andrae Palmer said:
Also does ColorSync handle the YUV conversion or just sets primaries, white point and gamma? If this works your way... Why haven't anyone on any platform been doing it this way before? Does ColorSync create a LUT for YUV color space? Obviously it can't send an sRGB monitor a YUV signal.
Click to expand...

Oops, missed this before.

YUV to RGB conversion is utterly trivial. They're not different color spaces, they're just different ways of encoding data. ColorSync is unnecessary for this conversion. All of FCP X's processing is in 32-bit float RGB anyway. If you're doing all of your processing with low-precision integer math, it might make sense to process YUV footage in a YUV color space to avoid rounding errors when converting. Once you move to 32-bit float processing, which gives you ~79 octillion possible RGB values for each pixel (no, really, I know that sounds made up, but it's true -- it's (2^32)^3; two possible values for each bit, 32 bits per color channel, three color channels per pixel) it's simpler to just do all of your processing in RGB. Resolve takes the same approach.
 
Chris am I understanding this right? You have been baited into defending FCPX as a primary tool for grading/finishing? Say it isn't so.
 
Tim Morten said:
A year at the top on the strength of a best-selling phone isn't a reason to deify Apple. They make mistakes like everyone else, and just like Microsoft, they won't stay at the top forever. Android is already outselling them in the phone space, and the tablet market is WAY too young to call. Apple remains at less than 10% of the PC market. iPod is their one uncontested win.

The Apple adoration runs thick with some, bordering on religious; Apple does some good work, and some not so good work. This thread is about a product that many of us consider to be in the latter category.

Personally, I think Apple is a better company when they are the underdog. Let's see what the future holds...
Click to expand...

Well, religious for some -- but now you can praise Apple purely on economics, and I never said they didn't mistakes. The Android "outselling them" reference is highly flawed, as I'm sure you know, of course with all the buy one get one free Android promos for $197 or $297 and availability on every major carrier .. OF COURSE they activate more phones right now. As Apple rolls out to more carriers, the gap will (and is) lessening -- and of course, smart phone "revenue" is a whole different story because it takes into account all the free and low cost Android deals. Out-activating them, yes? Outselling them - NO.

And while the tablet market is young, Apple has several HUGE advantages (market share, early adoption, best OS, most apps, and by far most important - marketing position as leader and "1st" - which the book Positioning by Ries and Trout shows makes it nearly impossible for anyone to overtake them.) Thus all signs point to iPod like dominance for many years to come.

Of course, despite being a "fan" (minus the boy) of Apple, I could tell you things I don't like about them:

- Still no 1080P in iTunes (WTF, though it's coming but late)
- No lossless or hi-res music
- Intentionally withholding features to create obsolescence (iPad 1 not having camera ... so that iPad 2 could)
- Seemingly lack of communication with market segments due to secrecy
- Overcharging on certain items (flimsy magsafe power adaptor - $80 - really??)

I don't "deify" them, but I feel I have a healthy and legitimate level of respect for them -- rooted in excellent products, numbers and yes my own subjective high level of satisfaction as a customer.

Anyway, like you said -- this thread isn't about all that, it's about FCPX.

As you said with the tablet market, it's too early to call :)
 
Last edited:
D Fuller said:
Well, perhaps you can't imagine it, but I've been in sessions where the differences were real and problematic for the grade. If a blue sky is banding, do you let it just band or do you manipulate the grade to avoid that? Perhaps more importantly, since darker colors are starved for bit depth in 8-bit color, how much do you care about banding in the shadows? And if you can't see the real result, how do you make decisions about how far you can push a grade? Chris, I don't believe for a minute that you believe this statement.
Click to expand...

Honestly, this kind of thing grows tiring. As the capabilities of low-cost hardware and software approaches have increased, some folks in this industry seem to have reacted to this by becoming obsessed with the increasingly marginal differences between what you can now do with off-the-shelf gear, and what you can do with expensive specialty gear. It's like the industry has collectively agreed not to believe in the concept of diminishing returns, because admitting to its existence would be bad for business.

Is 10-bit monitoring better? Yes. Should a shop charging outside clients for color grading have 10-bit monitoring? Probably yes, depending on how they present themselves and how much they charge. But let's get serious here. Watching the final result, it is ludicrously unlikely that you (or anyone) would be able to tell the difference between material graded while monitoring in 8-bit and material graded while monitoring in 10-bit. The difference is simply too minor to have a reliably detectable impact on the quality of the final product.

With respect to the issue of changing grading decisions based on 8-bit banding... visible banding doesn't occur at all in most shots. In shots where it does occur, unless you've pushed the image really far, you generally have a pretty good idea if you've got real banding or not, if you know what the source of the footage was. And do you even really want to grade material such that it looks fine on 10-bit screens, but has unacceptable banding on 8-bit screens? Even with features, more viewers will generally see the (8-bit) DVD/Blu-ray/streaming version than will see the film in theaters. Virtually everything other than theatrically released features will be shown exclusively from 8-bit formats. Frankly, you can make a compelling argument on this basis that it's actually advantageous to monitor in 8-bit, if you want your material to look as good as it can for the largest number of viewers.
 
Scott Crawley said:
Chris am I understanding this right? You have been baited into defending FCPX as a primary tool for grading/finishing? Say it isn't so.
Click to expand...

I'm mostly debunking a lot of nonsense about there being anything particularly unique about broadcast monitors.

But fundamentally... the raw quality of output from FCP X's rendering engine is as good as anything. The app supports 2K (and 4K, but that would probably get impractical). And I've screened enough ProRes 4444 in DI theaters to know it holds up just fine on big screens. I wouldn't finish a feature or serious broadcast project out of FCP X (with its built-in tools) because once you get used to grading with a control surface, and Resolve niceties like the 3D tracker, the nodal interface, shot grouping, etc. it's very hard to live without these things. Plus, of course, FCP X doesn't presently have native R3D support.

But the truth is, as much as some people are claiming FCP X is for "soccer moms", if someone shot a project in Log-C ProRes 4444 on Alexa, and had the patience to put up with the fact that FCP X's UI is not really well optimized for grading, they could probably produce something straight out of FCP X that (if they were competent) I don't think anyone could find any particular technical fault with. This is a reality that those of us who charge money for services like color grading need to come to terms with. Very soon, we will all be getting paid exclusively for our knowledge and skill, not because we have access to technical capabilities that our clients don't.
 
Chris,

Is it true that AVfoundation currently has no hooks for video? Essentially making no way third parties can accurately pull a video feed? It would be good if we limit the ColorSync conversation to accurate color monitoring vs accurate broadcast monitoring. Other than color accuracy there are other issues. I would love to hear from someone who has worked with ColorSync and profiling a srgb display to match a broadcast monitor. Why hasn't this workflow been used before? I'm still researching all the issues.
 
Andrae Palmer said:
Is it true that AVfoundation currently has no hooks for video? Essentially making no way third parties can accurately pull a video feed?
Click to expand...

QuickTime had a generalized architecture for handling video I/O devices and streams. There is speculation that FCP X currently lacks 'real' video output because AV Foundation does not yet have such an architecture, but anyone who can comment on the Mac version of AV Foundation (as opposed to the iOS version) is under NDA, because it's not a public framework pre-Lion, and the Lion developer docs are still under NDA until Lion's release.

If AV Foundation does presently lack such an architecture, I think it's likely Apple is working on this. They reportedly said at the London briefing that broadcast output via third-party I/O hardware is coming.
 
Chris... I love to be proven wrong because it means that I just learned something. If ColorSync can solve color accurate monitoring for rec709 via GPU that will be a huge step forward. I just sent an email to Flanders Scientific asking what issues we would have with rec709 monitoring via GPU utilizing ColorSync profiles. If this works properly I will definitely build a second workstation for "web color grading" using this method. I'll definitely save tons of money... That's for sure. I hope you are right. I'm not one to defend unnecessary expensive workflows. Right now I just want the facts before running to buy a spyder.
 
Andrae Palmer said:
Chris... I love to be proven wrong because it means that I just learned something. If ColorSync can solve color accurate monitoring for rec709 via GPU that will be a huge step forward. I just sent an email to Flanders Scientific asking what issues we would have with rec709 monitoring via GPU utilizing ColorSync profiles. If this works properly I will definitely build a second workstation for "web color grading" using this method. I'll definitely save tons of money... That's for sure. I hope you are right. I'm not one to defend unnecessary expensive workflows. Right now I just want the facts before running to buy a spyder.
Click to expand...

Not to impinge on anyone's integrity, but asking FSI if using a $600 Dell screen in place of one of their $5000 broadcast monitors is good enough is kind of like asking Red if shooting 2K ProRes on an Alexa instead of 5K raw on an Epic is good enough. You'll probably get an answer that, while technically accurate, is written from, let's say, a particular perspective. It's also possible that FSI has no particular expert knowledge with respect to this sort of color management workflow, because, after all, it's not widely used in the video world.

Also, note that as of right now, FCP X and, I believe, After Effects, are basically only video-related apps that can use ICC profiles in the right way to be useful for monitoring. So if you want to grade in Resolve or edit in anything other that FCP X, you still need 'real' monitoring.
 
Well these are guys that tried to dissuade me from buying their most expensive monitor. I think most people that have had relations with the company will tell you they are pretty honest. What are your thoughts on the other issues of broadcast monitoring such as proper frame rates and the ability to show interlaced footage? Or are you relegating this discussion to just web color accuracy?
 
Chris Kenny said:
You'll probably get an answer that, while technically accurate, is written from, let's say, a particular perspective.
Click to expand...

Every one of your posts is written with a certain perspective. We all know what it is too.
 
Andrae Palmer said:
Well these are guys that tried to dissuade me from buying their most expensive monitor. I think most people that have had relations with the company will tell you they are pretty honest. What are your thoughts on the other issues of broadcast monitoring such as proper frame rates and the ability to show interlaced footage?
Click to expand...

A fair number of GPUs can output at 24/25/50 Hz these days. As far as interlacing goes... modern display technologies are all natively progressive, so even broadcast monitors de-interlace internally before displaying. In theory equally good de-interlacing could be done in software for monitor display, but I have no idea if FCP X (or anything else) presently does this in a technically accurate way.

In general, though, if a 'real' broadcast video stream is going to be your final output, it's a good idea to have a monitor that can display one, scopes that can evaluate one, etc. But the idea that the broadcast stream, rather than the file on disk, is the final product, is probably going to slowly fade over the coming years.

Andrae Palmer said:
Or are you relegating this discussion to just web color accuracy?
Click to expand...

Well, general color accuracy. The best way to get "web color accuracy" (to the extent that it's possible) is just to calibrate a standard mid-range desktop monitor to sRGB at 2.2 gamma and not use any color management.
 
Chris Kenny said:
A fair number of GPUs can output at 24/25/50 Hz these days.
Click to expand...

The actual output from your graphics card is the monitor refresh rate you select in the driver. The number of unique frames displayed per second is up to the software you are using. To say "GPU's can output 24/25/50" suggests a misunderstanding of their architecture. GPU's can render anything the software chooses, from 1fps all the way to the monitor refresh.

- Tim
 
I think that there are a lot of holes in this ColorSync argument. It looks great on the surface, but...

1. the real accurate ColorSync math - the one Chris Kenny is using as an example where you are going from a RGB image to a CMYK printer image...

...is not implemented in realtime AFAIK.

I'm not saying it can't be done in realtime in the future. Actually, DaVinci Resolve, Smoke, Lustre, Avid DS etc do it today using a 3D LUT on the GPU. Which is pretty damn good.

But I haven't seen any evidence for this actually going on under the hood in FCP X.

2. There is a much lower-quality ColorSync mode which is just a couple of overall gamma curves for r, g, b.

This is what I believe FCP X is using on the fly.

This has some issues - for example, red on NTSC is a kinda unsaturated brick-like color compared to the vivid red of the current Apple monitors.

If you profile your monitor with a simple profile, ColorSync tries to desaturate the reds to fix the oversaturated primaries... BUT this then screws up your midtones. That was what I found in the past anyway.

So far in current tests, I am getting big differences between my (calibrated) DreamColor monitor in rec709 mode and the iMac display using Apple's own iMac profile.

I wanted to double-check today after profiling the iMac display with an i2 profiler - but FCP X is currently hanging on the splash screen saying "Loading Compressor Support" so I can't open the damn thing!

Chris - have you heard any mention of Apple actually using a real, proper 3D LUT?

Bruce Allen
www.boacinema.com
 
I'm still trying to grasp the differences between ICC profiles and 3D LUTs:

"There are other approaches to color management besides using ICC profiles. This is partly due to history and partly because of other needs than the ICC standard covers. The film and broadcasting industries make use of many of the same concepts, but they more frequently rely on boutique solutions. The film industry, for instance, often uses 3D LUTs (lookup table) to represent a complete color transformation. At the consumer level, color management currently applies more to still images than video, in which color management is still in its infancy.[1]" ----- http://en.wikipedia.org/wiki/Color_management

How do ICC profiles work in After Effects?
 
You must log in or register to reply here.
Back
Top