PDA

View Full Version : 32bit linear in After Effects


beenyweenies
03-20-2006, 06:12 AM
Just curious how many of you guys have played with 32bit linear yet. So far my experience has been an outstanding difference in overall quality and film-like appearance. This includes simple things like motion graphics, all the way down to color correction and image manipulation. Preview speeds have dropped somewhat on my dual-core PC system, but not enough to warrant ditching the process altogether.

I really notice the difference in things like motion blurs, transfer modes etc. I must say, I am totally impressed and look forward to ditching 8-bit. Anyone else using it yet?

Vympel
03-20-2006, 08:21 AM
Many effects is more beautiful using 32 bits, but not of all effects in AE is compatible with 32bits.
check this links have tips to work using 32 bits in AE

http://www.fxguide.com/article336.html

http://prolost.blogspot.com/ (good link tons of information over AE)

Mylenium
03-20-2006, 11:18 AM
Just curious how many of you guys have played with 32bit linear yet. So far my experience has been an outstanding difference in overall quality and film-like appearance. This includes simple things like motion graphics, all the way down to color correction and image manipulation. Preview speeds have dropped somewhat on my dual-core PC system, but not enough to warrant ditching the process altogether.

I really notice the difference in things like motion blurs, transfer modes etc. I must say, I am totally impressed and look forward to ditching 8-bit. Anyone else using it yet?

I think you're overstating the matter somewhat. You should see some difference from 8 bpc to 16 bpc for mograph stuff, but it is doubtful that you can justify using 32 bpc for what little additional benefit it may bring there especially when the final product is just a DVD that due to compression and limitations in color space is more like 7 bpc and not even true 8 bpc. The only exception would be color corrections of the very heavy type which will give less clipping but then again Color Finesse always has already worked in 32 bpc internally, alleviating the need to use workaroundish stacks of standard AE effects.

Of course the matter is entirely different for HDR stuff, but I don't see why you're getting this "film-like" appearance just from expanded color space. Maybe it's all in your mind?

BTW, several of the blendmodes still only calculate in 8 bpc/ 16 bpc even in a 32 bpc project, so it kinda is impossible that you would see any difference there.

Mylenium

Vympel
03-20-2006, 02:38 PM
Beyond of mylenium says, using HDR (32bist FP) can cause some types problems in specially with negative values

beenyweenies
03-20-2006, 06:35 PM
I think you're overstating the matter somewhat. You should see some difference from 8 bpc to 16 bpc for mograph stuff, but it is doubtful that you can justify using 32 bpc for what little additional benefit it may bring there especially when the final product is just a DVD that due to compression and limitations in color space is more like 7 bpc and not even true 8 bpc.

I don't agree. You are correct that output might ultimately be shown in 8bit, but output format has absolutely nothing to do with the development process. It's not as if you lose the results of working in 32bit because the output is 8bit. A good analogy might be saying there is no point in using HDR in 3D work because TVs can't display true HDR information. The results are embedded in your work, so it matters not what format they are shown in.

For example, when using effects that can handle 32bit (or using the available workarounds to work in 32bit>8bit>32bit) there is a much more filmic appearance. A good example might be blurs. Normally these look relatively dull, mushy and don't blend with underlying layers properly (dark halos and other unnatural issues), especially when compared to what is seen in real-world film blurs. In 32bit mode, when using the correct workflow, these look much more natural and "correct." To me, the goal is getting graphics and effects to mimick what people's eyes are used to seeing both in the real world and on the big screen.

...but I don't see why you're getting this "film-like" appearance just from expanded color space. Maybe it's all in your mind?
BTW, several of the blendmodes still only calculate in 8 bpc/ 16 bpc even in a 32 bpc project, so it kinda is impossible that you would see any difference there.


I wasn't implying that simply turning a comp 32bit will give any added quality or appearance, even though there can actually be a quality boost when it comes to aliasing and other issues. It's all about the results of doing certain operations that never look quite right in 8/16bit. Some transfer modes don't suffer anywhere near as bad from blow out/crushing in 32bit mode, and again mimick real-world film looks much better. But it requires the correct workflow to see these results. If you drop a layer into Add Mode (32bit's equivilent to Screen Mode) the results are very apparent, and much improved, compared to using Screen Mode in 8/16bit. I have a still that proves this point which I will upload tomorrow when I am at my work machine.
You are right that you can't use everything in true 32bit, such as some transfer modes etc. but the whole point is that you no longer NEED to, since most of the old ways are based on faking, and correcting for the lack of quality and film-like appearance of 8/16bit.

By the way, when I say film-like appearance, I don't mean mimicking a particular stock, I just mean duplicating real-world visual phenomena/reality in a computer-based world.

The only exception would be color corrections of the very heavy type which will give less clipping but then again Color Finesse always has already worked in 32 bpc internally, alleviating the need to use workaroundish stacks of standard AE effects.

Color correction gets a nice boost in latitude and range in 32bit, and while you are correct that Color Finnesse works OK, I don't know any professionals who actually use it. Not to say it is an amateur tool, it works great. But launching a seperate application every time you need to tweak a setting is a bit annoying, heavy-handed and adds to development time. In the end, the added render time is comparable to just working in true 32bit mode anyway. Time saved, same or better results, simpler pipeline. This improved (IMO) workflow allows me to use standard Levels to do all color correction, without losing any quality, and with only a slight modification to workflow which is second nature after a few uses.

A previous poster mentioned that not all effects are 32bit. Well, actually most aren't 32bit but there are adequate workarounds that I have found work just fine. While these workarounds result in less than perfect 32bit quality, they are still much better than 8bit or 16bit.

Needless to say this process is still in its infancy and there will be much debate over the merits of working in 32bit. My stance is simple - in this crowded field of motion graphics and effects work, an edge is needed. People willing to take on the complexities of working in 32bit WILL see positive results if done correctly, and the customer will too.

Blur1
03-22-2006, 11:29 PM
I have to say I agree with you Beenyweenies (great name BTW). I was initially closed off to the idea of AE getting float because I thought the programmers should focus on other areas first but they've done it and it's great to have it in there.
Mylenium, you may want to do some trawling on the prolost site to see why hdr comping is good for 8 bit delivery. It simplifies the workflow in many situations and delivers nicer results. There's a great chapter about it in the Mark Christensen AE book. Also keep in mind that a projection positive of a film down at the multiplex is probably only about 10 linear bits anyway, so the delivery medium is generally always going to be a subset of the working dynamic range.

Mylenium
03-23-2006, 05:48 AM
Mylenium, you may want to do some trawling on the prolost site to see why hdr comping is good for 8 bit delivery. It simplifies the workflow in many situations and delivers nicer results. There's a great chapter about it in the Mark Christensen AE book. Also keep in mind that a projection positive of a film down at the multiplex is probably only about 10 linear bits anyway, so the delivery medium is generally always going to be a subset of the working dynamic range.

I know Stu's website (as I do know him a tiny bit) but I don't completely agree on his views in some areas nor do I share beenyweenies assesment of 32 bit being great for motion graphics in general - that what my post was all about, not condemmning 32 bpc as such or denying the need for it on some level.

I dunno what you do for a living, but I'm constantly faced with doing 10000 frames trailers and animations in only one weeks time for tradeshows and the like and neither is 32 bpc doing anything to make them look better nor does the more complex handling due to larger file sizes, memory consumption and longer render times do any good. Similarly there is no point in blowing up a DigiBeta shot to 32 bit and pretend it's going to gain any quality just by that. On the rare occasions when I had the chance to work with film stock or HD footage, I used 16 bpc. I haven't had that type of project in a while and haven't tested how AE holds up there, but it sure is one thing to try and look forward to.

I also need to point out that motion graphics to me means just plain old SD stuff going to DVD and broadcast about 80% of the time and as you may or may not know, the HD ball only get's rolling here in Europe (and very slowly at that) which means that 32 bpc so far has little or no relevance for most day-to-day tasks in most companies in this area. Granted, it will all change and we'll all have to adapt, but at the moment it's a very dry and theoretical topic from my POV.

In addition to that I think that we all can be honest here: There is no need to pretend that what you all claim to be a matter of quality and honor here, holds up to scrutiny in your everyday workflows, even if you work with HD and film stock. We all have times when we are so in the crunch that we forego any of the "fancies" if it means our stuff will render fast enough to meet the deadline and as long as it passes quality control (or the clients judgement).

Mylenium

beenyweenies
03-23-2006, 09:26 PM
Mylenium,
In the interest of debate and bouncing this topic around, I'd like to respond to your post. Not to contradict you because certainly we all have our own workflows and ideas, But I think you are understating things, or maybe even missing the point:

I'm constantly faced with doing 10000 frames trailers and animations in only one weeks time for tradeshows and the like and neither is 32 bpc doing anything to make them look better nor does the more complex handling due to larger file sizes, memory consumption and longer render times do any good.

First of all it is simply NOT true that using 32bit will do nothing to make your work look better, if you are using the correct workflow. By this I mean you can't just toggle on 32bit and expect to see dramatic results, there is a specific workflow involved to get things working properly which Stu outlines in great detail on his blog. (http://prolost.blogspot.com) How valuable or necessary the improvement is may be a matter of need, taste and budget - but you can't say 32bit does nothing to improve quality. Another thing is that using 32bit should not increase your file sizes at all - either project files or output files. If you are importing HDR imagery for use with 32bit comps then yes storage needs will be larger to accomodate those files, but I haven't even used any external HDR images yet and the results are still worthwhile to me. Here's a few samples that I threw together VERY quickly to make my point:

http://www.splitvisiondigital.com/files/hdr_addCompare.jpg
This is a comparison of blending in 8bit vs. 32bit. While the difference is subtle and this example isn't as dramatic as many I've seen, you can see how detail is retained much better in a 32bit world.

Next up:
http://www.splitvisiondigital.com/files/hdr_blurCompare.jpg
Again, the results are not as dramatic in this example, but you can tell that simple blurs (this is using Fast Blur) do not suffer from the black halo, unnatural composite that 8bit does. A better example might have been text with motion blur flying quickly across the screen - in such cases the differences are very obvious to anyone with an eye for detail.

While it's true that memory usage and render times are expanded slightly, it's really not that bad unless you are on an already underpowered machine. Yesterday I was playing around with 32bit comps on my girlfriend's 5 year-old 800mhz HP Pavilion with less than 300MB RAM and it worked just fine.

I also need to point out that motion graphics to me means just plain old SD stuff going to DVD and broadcast about 80% of the time and as you may or may not know, the HD ball only get's rolling here in Europe (and very slowly at that) which means that 32 bpc so far has little or no relevance for most day-to-day tasks in most companies in this area.

I do a pretty mixed array of work myself, including a lot of tradeshow graphics, so I know timeframes and output formats can be limiting. But again I think you are selling 32bit short. You DO NOT need to be outputting to a high-resolution format like HD or Digi-Beta to see the results of 32bit comping. If I am doing some 3D work and I use an HDR image for image-based lighting, will I lose the visual quality of that by outputting to 8 bit? As you can see in the images I posted above (even though they aren't the best or most compelling examples) there is a difference even when outputting to simple Jpegs, so why would 8 bit output be any different?
Also, thus far I have not experienced any dramatic increase in design time simply from using 32bit. You can easily set up a 32bit comp workspace with all the necessary ingredients already in place, and save this out as a template project. Fire it up and work as usual with very little additional workflow hassle needed.

There is no need to pretend that what you all claim to be a matter of quality and honor here, holds up to scrutiny in your everyday workflows, even if you work with HD and film stock. We all have times when we are so in the crunch that we forego any of the "fancies" if it means our stuff will render fast enough to meet the deadline and as long as it passes quality control (or the clients judgement).

Certainly there are situations where 32bit would not make sense - however, the added render times you speak of simply don't seem to be an issue for me. On my machine render time difference between 8bit and 32bit were not severe, and in some situations were no different at all simply because I could abandon messy comp/filter structures used to circumvent the quality problems of 8bit. For example, when doing critical color correction using 32bit workflow instead of an app like Color Finnesse the results are similar or better, render times are the same or better, and the time savings in workflow alone makes it all worthwhile.

My intention is not to force 32bit on anyone or to claim that it will somehow make you a better artist. Simply put, it is a significant leap forward that, much like HDR in general, will revolutionize AE work once it is more widely adopted and the benefits become more obvious. As always I am open to counter-points or input disproving my claims.

Blur1
03-23-2006, 09:58 PM
Well stated, and it has to be said that even if you were comping with 8 bit jpegs taken with your phone, you can get a nicer result in float space, with less operations. Unless you are working with 10 bit HD with a special Cineon-type LUT that can be spread out into a larger dynamic range in the compositing process, HD formats are going to be the same dynamic range as Digibeta, ie, 10 linear bits so there is no real exoticism there except for the increased spatial resolution.

I will concede that you could probably just work in a gamma 1.0 comp in 16 bit and this will add the "linear light" comping which for example improves blending of blurs and chromatically opposite colours. But you will forfeit control of the extended dynamic range of intermediate operations.
Also I agree Mylenium that it is not relevant for every job but as I become more comfortable with the colour management issues, and as Adobe starts to improve this, I will use it more. I would like to see bit depth assigned on a per-comp basis.

Vympel
03-23-2006, 11:31 PM
I will concede that you could probably just work in a gamma 1.0 comp in 16 bit and this will add the "linear light" comping which for example improves blending of blurs and chromatically opposite colours. But you will forfeit control of the extended dynamic range of intermediate operations.
Also I agree Mylenium that it is not relevant for every job but as I become more comfortable with the colour management issues, and as Adobe starts to improve this, I will use it more. I would like to see bit depth assigned on a per-comp basis.

I think the ideal is control the bit depth on a per-"layer" basis like others applications (Combustion, Shake...), in these applications you can use the bit Depth converter and promote and depromote the footage in any part of composition, allowing you to promote the bit depth only is necessary like in color conversions/correction. The AE needs of systems to control the visualization of files with high bit-depth or in logarithmic space (LUT views)

beenyweenies
03-24-2006, 01:46 AM
I think the ideal is control the bit depth on a per-"layer" basis like others applications (Combustion, Shake...), in these applications you can use the bit Depth converter and promote and depromote the footage in any part of composition, allowing you to promote the bit depth only is necessary like in color conversions/correction. The AE needs of systems to control the visualization of files with high bit-depth or in logarithmic space (LUT views)

Agreed. For now let's all keep in mind that 32bit AE is in its infancy and will take some time to get there. They themselves concede that this is a baby step toward true 32 bit workflow, hence the lack of a full set of native 32bit filters, etc. Until then, there are sufficient workarounds, and a sufficient set of filters, to keep most people happy.

Mylenium
03-24-2006, 05:53 AM
While it's true that memory usage and render times are expanded slightly, it's really not that bad unless you are on an already underpowered machine. Yesterday I was playing around with 32bit comps on my girlfriend's 5 year-old 800mhz HP Pavilion with less than 300MB RAM and it worked just fine.
...
Certainly there are situations where 32bit would not make sense - however, the added render times you speak of simply don't seem to be an issue for me. On my machine render time difference between 8bit and 32bit were not severe, and in some situations were no different at all simply because I could abandon messy comp/filter structures used to circumvent the quality problems of 8bit.

That's the point I would argue the most. Fact: AE handles 32 bpc at least 4 times slower even sans effects than it handles 8 bpc/ 16 bpc. With several layers stacked one upon another this can quite add up and I don't think it's making 32 bpc more sexy.

As for the rest - well, you're still sounding like an 32 bpc evangelist. Not that I don't see your points (as well as I do see Stu's points and think they are very valid), but it doesn't mean I agree to them. Your initial post was pertaining to the use of 32 bpc for MoGraph and with all due respect, that is the one area of CG where I think all this color space expansion talk is overrated by far and least relevant.

Sticking to your examples, let's assume you put together the most beautiful tradeshow trailer and put in all the technical finesse and knowledge you have. So what happens?

a) The finished product goes to a DVD or tape, already effectively reducing color fidelity as discussed earlier and quite drastically, I might add (4:2:0 for DVD, 4:1:1 for NTSC-DV!). In addition to that you're having a hard time getting artefact-free compression, because 32 bpc is "too good" and 8 bpc/16 bpc with their "wrong" blend modes and added bluriness work much better for MPEG-II or DV.

b) Your stuff is displayed on burnt-out plasma displays, poorly calibrated cube walls, weak beamers or LED grids (6 bpc at best!).

c) Your client doesn't care, as long as it looks good enough and is cheap. He may even be so impertinent as to elect to run his presentations from within PowerPoint or his DVD player software on his notebook.

So in the not so rare realworld situations all claims of sticking to quality and working in the best interest of the client are pointless, pretentious and ridiculous because only in 1 out of 30 productions the client has even the least aprreciation for such matters. That is all I'm saying and one of the main reasons I will be sticking with 8 bpc/ 16 bpc for a lot of stuff. It's a business decision, not an artistic one.

Mylenium

beenyweenies
03-24-2006, 07:53 AM
So in the not so rare realworld situations all claims of sticking to quality and working in the best interest of the client are pointless, pretentious and ridiculous because only in 1 out of 30 productions the client has even the least aprreciation for such matters. That is all I'm saying and one of the main reasons I will be sticking with 8 bpc/ 16 bpc for a lot of stuff. It's a business decision, not an artistic one.

Mylenium

Point taken, and on this we are somewhat in agreement. For the situations you outlined (output to DV tape) it would be overkill using 32bit to get more realistic blend modes or blurs when the end display is ultra low resolution and heavily crushed. But as with all things it is entirely dependent on your work environment and needs. Personally I am used to working more in the Digi-Beta / HDCam / Film world where 1 out of 30 clients don't care, rather than the opposite scenario you illustrate. In these situations quality is scrutinized with a magnifying glass and so 32bit can make much more sense - it is these scenarios that my evangelism is aimed.

As for 32bit being useless for MoGraph, I still don't agree. In fact, I think this is one area that could use an upgrade given all the highly computer-generated looking graphics out there. My personal take is that a new trend could emerge as a result of 32bit, particularly a more natural, photographic approach rather than candy-colored, heavy graphics. Style aside, I do think there is benefit to retaining more detail, having less blown out and crushed areas, and having less banding, particularly in HD presentations.

Mylenium
03-24-2006, 08:28 AM
As for 32bit being useless for MoGraph, I still don't agree. In fact, I think this is one area that could use an upgrade given all the highly computer-generated looking graphics out there. My personal take is that a new trend could emerge as a result of 32bit, particularly a more natural, photographic approach rather than candy-colored, heavy graphics. Style aside, I do think there is benefit to retaining more detail, having less blown out and crushed areas, and having less banding, particularly in HD presentations.

That would be desireable, but that's not going to happen. You know, you can see so much crappy on-air graphics on TV everyday as you can see shabby ads and marketing stuff even from big companies who supposedly care for quality and have the money to make it a point of their corporate image, but it hasn't compelled anyone of the people responsible to really make use of technical advancements. I think from our end as graphics artists it's the least problem. The best you can hope for is that technical need (e.g. greater bit-depth to avoid banding in HD) will push others to adapt to the new possibilities.

In addition to that I think you're dismissing the cultural background always relevant for that type of stuff all too easily. Natural and photographic may be what you and you're clients want, but not others. Quite frankly I don't necessarily consider a lot of what is done in the states as particularly stylish or artsy (often too flashy, overdone and with a touch of insane grandeur; too warm colors) and the other way around you probably think a lot of our reduced European designs are dull and boring. Chinese just love reds and gold and as much as it sounds like a sterotype - Indians love floral designs in pastels. Same goes for people at different ages. You can't win over a 15 year old adolescent with your natural flowers, when he plays Halflife all the time and actually loves this synthetic, technical look and style. To some degree that alone goes against using 32 bpc for everything and anything - it may actually be quite welcome to get clipping in 8 bpc if it helps to produce a videogame look.

Mylenium

SirRender
03-24-2006, 07:14 PM
I just want to say that I've enjoyed this thread very much. Thank you for bringing up the subject beenyweenies and thank you Mylenium your thoughts on the matter.

I too am a early adapter of 32bit linear workflow in AfterEffects and pre-AE 7 using eLin.

I do a lot of HD and 2k renders for tradeshow graphics and most of my work is still done is 8bit. Most clients (specially for this type of work) are going to always want what's good enough and the cheapest price.

IMO, that's not point of using a 32bit linear workflow. Weather it makes sense for you depends on a lot of different issues, and lets not even get started on personal design/graphics tastes because thats a whole different discussion.

I use 32bit workflow when its appropriate for me. I love that I have the option to do this in AE 7 now, weather its for motion graphics or television/film compositing. Clients will probably not notice it, but as an artist looking to constantly improve my work, I will notice. That's basically what it comes down to for me. Do what you need to do to make your deadline, and if working in 32bit works in your pipeline, then more power to you. If not, then don't, but don't tell me that's its a waste of time because the subtleties are not worth it for your pipeline.

I would love to continue discussing 32bit workflow in AE, specially how to incorporate openEXR 3d renders (non-gamma encoded) into the workflow. Maybe in a different thread?

Mylenium
03-24-2006, 07:20 PM
I would love to continue discussing 32bit workflow in AE, specially how to incorporate openEXR 3d renders (non-gamma encoded) into the workflow. Maybe in a different thread?

We could do that. I think there's always a default Gamma in any EXR file, it's just that every program uses another value. Will have to read the specs again.

Mylenium

Vympel
03-24-2006, 08:01 PM
The Gamma of EXR files is 1 (other formats have a internal gamma of 2,2,sRGB), to view this file correct (or other files with the inverse gamma of 2.2 applied, the macros of prolost to Shake and Fusion make this, I believe)you need apply a view correction of 2.2 gamma

SirRender
03-24-2006, 10:25 PM
Yes, Vympel. I believe your are correct. Does anybody know how to set up this view correction when rendering EXR files out of Maya Mental Ray?

I know that I need to change my frame buffer to RGBA [Float] 4x32 Bit in order to get the 32bit float info, but do I also need to setup this gamma view correction in order to get a clean 32bit float EXR image that is not windows pc 2.2 monitor gama encoded?
From listening to Stu's fxguide podcast he mentions how a lot of people are rendering and using gamma encoded EXR, which is NOT what you really want.

Anybody doing EXR renders out there?

Vympel
03-24-2006, 10:57 PM
The openEXR for default have no gamma correction applied, but you can apply accidentally a gamma correction when you setting the images options or render the images (in this link have some ideas to work in AE with 32bits, like in others app http://prolost.blogspot.com/2006/02...ng-icy-sea.html ). This link have a info about use of HDR files in the mental ray in XSI (http://www.xsi-blog.com/?cat=6)

Blur1
03-25-2006, 12:07 AM
I think the ideal is control the bit depth on a per-"layer" basis like others applications (Combustion, Shake...), in these applications you can use the bit Depth converter and promote and depromote the footage in any part of composition, allowing you to promote the bit depth only is necessary like in color conversions/correction. The AE needs of systems to control the visualization of files with high bit-depth or in logarithmic space (LUT views)

Yes, you're right, but I didn't want the AE coders to get ahead of themselves ;0

beenyweenies
03-25-2006, 12:53 AM
I use 32bit workflow when its appropriate for me. I love that I have the option to do this in AE 7 now, weather its for motion graphics or television/film compositing. Clients will probably not notice it, but as an artist looking to constantly improve my work, I will notice. That's basically what it comes down to for me.

Thanks for the comments, and I agree with you 100%. One thing we're probably all facing is tight deadlines, and apparently many people sense indifference from clients when it comes to quality, so it's easy to see why they may hesitate to add time and complexity to their workflows. Here's my take on that.

The client is typically NOT an artist - in fact the typical client is a "suit" or other action-oriented person who's primary interest is getting the message across without breaking the bank. This includes producers, many of whom are glorified bean-counters with little or no actual artistic experience. Regardless, their role in the creation process is simply to get. it. done. Therefore, their demands with regard to quality should serve as the floor, not the ceiling, of quality we expect from ourselves. If this were not true, there would be no innovation - plugins such as 3d-stroke would not be cost effective because the client is not demanding such things - at least not until they are in frequent use, and generate popular demand. It's called thought-leadership. No matter how foolishly idealistic it may seem, I see myself as an artist, not a cog in a sausage factory. To this end, I like to push myself and the technology to acheive better each time I sit down to do a job. If this means taking advantage of a quality boost, so be it. But I am supremely confident that new plugins, techniques and other exciting developments will emerge from this 32bit workflow. You can already see such gains in apps such as Shake.

HD is moving FAST, and it's only a matter of time (here in the states, anyway) before digital TV is a legally mandated requirement. Therefore, the quality issues Mylenium addressed may not be such an issue for long, and as they say, jump aboard or get left behind.

beenyweenies
03-25-2006, 01:00 AM
The Gamma of EXR files is 1 (other formats have a internal gamma of 2,2,sRGB), to view this file correct (or other files with the inverse gamma of 2.2 applied, the macros of prolost to Shake and Fusion make this, I believe)you need apply a view correction of 2.2 gamma

I have yet to get into using openEXR so any discussion on this will be tasty, indeed! So a gamma of 1 is what makes it linear, correct? My impression is that from there, anything done is merely for the sake of viewing it correctly, but not applied permanently to the file itself. For example, you would NOT want to apply any gamma "viewing" correction to your renders, merely wait until it's in your compositing app and THEN apply a LUT of sorts?

I use maya, but have never touched OpenEXR. Where's a good place to start? Any of the sites have good threads on this topic, such as highend3D or CGTalk?

amv256
03-27-2006, 05:37 PM
This is a very informative thread so far. I have a few kinda newbie questions for the 32-bit guys here. I've done a lot of AE work for both broadcast and more private venues like trade shows myself, but I've only never done anything in 16 BPC, much less 32:

1) Since obviously our monitors are still all 8 BPC, we're not actually SEEING 16 BPC and 32 BPC images, we're only seeing the results of their presence in the workflow, correct? I just want to make sure there isn't some sudden surge in > 8 BPC monitors out there that I haven't heard of. :)

2) Just for my own edification, can someone outline a couple specific instances where using 16 or 32 BPC is either very important, or downright necessary? Like I said, I've done all of my mograf/3D work in broadcast and similar venues so far; I've never done feature film, nor do I work with actual video footage much to begin with (99% mograf and 3D). So with that in mind, what are some cases where 16/32 is more common? I'm gathering that film work seems to necessitate these modes. Any other good examples?

It'd be great if I could get a better understanding of these modes and their place in the mograf/video/tv/film industries to at least start investigating it more clearly.

Thanks!

Mylenium
03-28-2006, 04:36 AM
1) Since obviously our monitors are still all 8 BPC, we're not actually SEEING 16 BPC and 32 BPC images, we're only seeing the results of their presence in the workflow, correct? I just want to make sure there isn't some sudden surge in > 8 BPC monitors out there that I haven't heard of. :)

Not entirely true. Extremely good (and expensive) CRT monitors can go up to 12 bit fidelity but even with the best phosphors this will not be spread equally throughout the spectrum. Still, you are right about seeing the presence of higher bitrates even on the most shabby displays simply because of less noticeable dithering/ banding. Which leads us to your next point.


2) Just for my own edification, can someone outline a couple specific instances where using 16 or 32 BPC is either very important, or downright necessary? Like I said, I've done all of my mograf/3D work in broadcast and similar venues so far; I've never done feature film, nor do I work with actual video footage much to begin with (99% mograf and 3D). So with that in mind, what are some cases where 16/32 is more common? I'm gathering that film work seems to necessitate these modes. Any other good examples?

Technically, HD stuff requires at least 10 bpc (a 8 bit gradient has only 1024 distinct levels of brightness which is'nt enough to cover the entire width of a HD image even at the smaller 720 size). Other than that, higher bit depth is advisable for color corrections, even of the minor sort. I'm using 16 bpc most of the time simply because it minimizes clipping if using e.g. the Hue/ Saturation effect. In a 3D workflow you may want to consider the difference between using the Levels effect and the new Exposure effect when adjusting HDR and EXR files - they are both completely different things and the latter beats the first every time. So far that's the only point where I'm really willing to get into 32 bpc. Film does not necessitate using 32 bpc, but it surely helps.

Mylenium

Blur1
03-28-2006, 05:34 AM
Technically, HD stuff requires at least 10 bpc (a 8 bit gradient has only 1024 distinct levels of brightness which is'nt enough to cover the entire width of a HD image even at the smaller 720 size). Mylenium

8 bit is 256 levels, 10 bit is 1024 levels of dynamic range. Although you are drawing a parallel between spatial resolution and dynamic range, this is misleading as far as real world practice is concerned. Even DVCProHD (the Varicam format) is only 8 bit.

HD does not "require" a higher working bit depth because the only guarantee with your source HD footage is that it has more spatial resolution.

Other than that, higher bit depth is advisable for color corrections, even of the minor sort. I'm using 16 bpc most of the time simply because it minimizes clipping if using e.g. the Hue/ Saturation effect. Mylenium

You will still get clipping of overbrites with 16 bit because it's an integer format. You must be talking about blending of adjacent values.


In a 3D workflow you may want to consider the difference between using the Levels effect and the new Exposure effect when adjusting HDR and EXR files - they are both completely different things and the latter beats the first every time.
Mylenium

It is possible to reconfigure the Levels effect to behave like the Exposure effect by using an expression attached to output white(see the eLin docs), so I disagree that the Exposure effect beats Levels "every time". The Exposure effect is a subset of the capabilities of Levels so you will get exactly the same result either way as long as you uncheck the "clip values" checkbox in Levels when working in float.

A good reference for someone wanting to start with 32bpc in AE 7 would be the gurulounge podcast on this very subject:

http://www.totaltraining.net/gurulounge/

beenyweenies
03-28-2006, 06:48 AM
Just for my own edification, can someone outline a couple specific instances where using 16 or 32 BPC is either very important, or downright necessary? Like I said, I've done all of my mograf/3D work in broadcast and similar venues so far; I've never done feature film, nor do I work with actual video footage much to begin with (99% mograf and 3D).
Thanks!

The primary situations under which moving up to 16/32bit makes sense are color correction, gradients, visual effects (including greenscreen work) work with a lot of motion blurs and maybe a few others. Color correction reasoning in explained in other posts, gradients do not band anywhere near as bad in 16bit+, visual effects get more info to work with and usually turn out better, and the last, motion blurs, applies to 32bit. If set up properly, you can get really nice blurs (especially motion blur) in 32bit mode - something much more photographic and natural in the way they blend.

I would say the above benefits apply no matter what the final medium (film vs. trade show video) because they make your overall work better. As long as you can do it within your budget, better is good. ;)

Vympel
03-28-2006, 03:07 PM
ther than that, higher bit depth is advisable for color corrections, even of the minor sort. I'm using 16 bpc most of the time simply because it minimizes clipping if using e.g. the Hue/ Saturation effect. Mylenium

Only band artefacts is reduced, clipping continue occuring (limited from 0 to 1) See this link

http://www.digitalgypsy.com/vfxlog/archives/tip_of_the_week/index.php



Originally Posted by amv256
1) Since obviously our monitors are still all 8 BPC, we're not actually SEEING 16 BPC and 32 BPC images, we're only seeing the results of their presence in the workflow, correct? I just want to make sure there isn't some sudden surge in > 8 BPC monitors out there that I haven't heard of.


Not entirely true. Extremely good (and expensive) CRT monitors can go up to 12 bit fidelity but even with the best phosphors this will not be spread equally throughout the spectrum. Still, you are right about seeing the presence of higher bitrates even on the most shabby displays simply because of less noticeable dithering/ banding. Which leads us to your next point.

You can use View LUT to control the "expose image" of files with high bit depth.

Good link about HDR images

http://www.fxguide.com/article268.html

CGTalk Moderation
03-28-2006, 03:07 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.