Flatting the flat look

UPDATE MAY 2011: Canon engineers, along some specialists from Hollywood, developed the Technicolor CineStyle profile. It’s even flatter than superflat or extraflat profiles, with better color accuracy and dynamic range in the shadows. Don’t waste your time with any other profile, just get that one.

As you probably know, it’s important to shoot “flat” with your video camera, in order to help color grading in post, especially if you’re after the “film look”. Since I got the Canon 5D MarkII, I made sure I shoot as flat as possible: with the “Neutral” color setting, modified to have the contrast/saturation/sharpness settings on minimum, and its tint on +1.

So far, I’ve only shot two videos with the 5D, and I was not happy with the visual result. Yes, its picture quality is amazing for the price, but as the occasional filmmaker & colorist that I am, I need the kind of look directly out of the camera that I could get with a film camera, or the RED One. And the 5D, with tricked out color settings, it would still not give me what I wanted: the videos came out over-saturated, and over-contrasty for my taste. Sure, the videos were magnitudes less contrasty/saturated than when using the “auto” color modes, but they were still not what I wanted.

I tweeted about it yesterday, and some people suggested I try the Canon Picture Style Editor, which lets you edit these parameters even more. So I downloaded this famous package, which includes a Panalog-like curve (which I didn’t like), Marvel’s EX1-like Cine curve, one called “superflat”, and a pseudo-Velvia one.

Well, I’m still not happy with the results. None of these downloadable styles are what I wanted exactly, and to make the matter worse, the curve utility inside the Picture Style Editor sucks goats: you can’t move the two edges of the curve. Photoshop’s curve dialog can do it, but Canon’s utility can’t. Because of that, it’s impossible to get more detail in the dark places directly out of the camera. You see, whatever you can do IN-camera, is MUCH more desirable than doing it in post processing. IN-camera processing is higher quality, so what you get out of it is purer, and doesn’t bring out as much the h.264 artifacts when lowering contrast in post. But without a curve dialog that let’s me do more, I can’t tell the camera to shoot that way. To be fair, this feature didn’t need to exist in the past, because photographers don’t care about it, but filmmakers do. Now that dSLRs can shoot video, hopefully a better curve dialog will be implemented in the future by Canon.

So, I had to do with what I had. I edited Neil Stubbings’ “superflat” style, and created a new one called “ExtraFlat”. My version uses the “Neutral” look as a base instead, and it’s a tiny bit less contrasty, but a lot less saturated, and it doesn’t have the “red face” attribute of the video look. Of course, we should not forget that Canon uses extra processing when sizes down the sensor image to 1080p, but that’s a kind of processing we can’t control.


The ExtraFlat picture style was used to shoot this music video. Color grading in post was minimal.

You can download the ExtraFlat style for your Canon vDSLR camera here. Instructions on how to upload it to your camera after unzipping it, here. Make sure the “EOS Utility” is installed on your computer. Check the ExtraFlat style compared to the rest.

Now, please don’t start commenting again about how you prefer the “standard” contrasty/saturated look. I don’t care if it looks better as a FINAL still picture. Don’t think of this frame grab as the end result. Video footage of any artistic work HAS to go through color grading, and for that, you need a FLAT look to work on.

Look at how the pros do it. Check this 2k frame grab, directly out of the RED One camera (ungraded). Notice how it’s extremely low-saturation, low-contrast, and the people’s skin is almost PINK-GREY and not red as the consumer camcorders do it (check this HV20 frame for reference). With a Canon consumer camcorder, even if you use Cinemode+custom color settings, it’ll still look red-ish, compared to what the RED ONE does. Panasonic consumer HD cams are way worse, since not only they don’t go as far in color settings, but their footage is processed to be very red by default.

With the ExtraFlat style I get almost what I want out of the 5D, but more dynamic range could be acquired in-camera if the curve dialog in the Picture Style Editor was better implemented. The camera CAN do it, we just don’t have a way to TELL it to do it right now.

Update: Shot a small video of me testing the flesh tones of the ExtraFlat today. It was flat and non-red alright! And it graded so nicely. Going through the various grading templates, it offered a very pleasant look, across the board.

27 Comments »

rhervag wrote on January 25th, 2010 at 1:50 AM PST:

you’re upset because you want the tool to do it your way while it might be smarter to embrace its strength and work around its limitations.
using the shoot flat + heavy grade workflow is not such a great idea on current DSLRs output.
if you want to play with grading get real and rent or buy a RED.

there is no way h264 will give you any latitude to do it right after the sensor gets subsampled to 8bit. you can flat out contrast and then cineform the heck of it, all the smoothness and subtle details are already lost for good before arriving on the CF card. period.

it is much better to commit to a look during pre-production and bake out a few key picture styles for use on location according to lighting conditions to preserve the maximum bandwidth with relevant data.
the more you correct such footage the worse it looks so how about getting it right in the first place with correct exposure and full bandwidth.
that will at least put the initial footage in the right ballpark, leaving minimal adjustment to the colorist and save veryone from embarrassing results.


This is the admin speaking...
Eugenia wrote on January 25th, 2010 at 2:02 AM PST:

>while it might be smarter to embrace its strength and work around its limitations.

This is exactly what I do by taking the time and making my own picture style (among other things, of course). I’m coming from a very geek background, so tweaking, and working around limitations, is my thing. You’re preaching to the choir… ;)

>you’re upset because you want the tool to do it your way

No, I’m not upset, you misunderstood. The curve dialog is not fully implemented, that’s true, but it’s not the end of the world. The ExtraFlat picture style does the job for the most part.

>if you want to play with grading get real and rent or buy a RED.

I have absolutely no desire to either buy or rent a RED. I DON’T NEED one. It’s too much trouble for what I do. It would require a new Mac, with FCS too, in order to work properly with it. And since I only have a slower PC+Vegas, that translates to time+money waste for me. Not to mention that I might need a crew for a RED too. So, no, the RED is definitely not what I need for the simple stuff I do.

I spent $2k and bought a camera that fits my needs in a multitude of ways, and I try to go around its limitations any way I can. I’m cool with that.

>are already lost for good before arriving on the CF card.

I don’t disagree with that, but I get what I paid for. For the $2k I paid for the 5D (second hand), I get a more pleasurable image than I would with a consumer camcorder at a similar price. So everything is relative. And besides, color grading a Cineform’ed 5D footage is not bad at all. It comes out very acceptable.

Is the 5D as good as a 4k RED? No. But I don’t need it to be either. I don’t need it to be perfect. I just want it to be good-enough. And these dSLR cameras are 90% there for the kind of stuff *I* need. For what I do, where I don’t charge a dime for my music video work, they do the job much better than the guy’s HDV camcorder next door.


rhervag wrote on January 25th, 2010 at 2:20 AM PST:

alright there but considering you have grown such a big influence on wannabe filmmakers that follow your blog and forum posts, you might want to also provide the other side of the coin so they know what to expect.
i mean it’s fine and dandy to play with magic bullet and all but it gets actually disappointing if pushed too much when projected on a big screen. outrageous banding and compression artefacts which look minimal on a 40″ LCD will jump at the viewers on a 52′ screen. negligeable contrast issues and gamma shifts are a concern on such bigger surface with grainy shades crawling like hell. trust me, i’ve just been there over the summer and had to reconsider such method from the start. getting the most optically through a proper picture style while recording for minimal fuss in post yields far better results. don’t shoot flat : shoot it right on the set.

obviousy for DVD & 720p web delivery, you can magic bulletise the hell out of the flattest underexposed footage and it will still look cool in most cases…


This is the admin speaking...
Eugenia wrote on January 25th, 2010 at 2:34 AM PST:

>will jump at the viewers on a 52′ screen.

I have a 50″ Pioneer plasma (one of the best TV models ever released), and it looks great. I have no issues whatsoever.

I don’t push color grading much for the videos I do. For the tests I do on my blog, I color grade the hell out of them, but not for the actual videos I publish. I’m more careful on those.

Oh, and I usually try to avoid Magic Bullet, and use the 3-way color corrector on Vegas directly. I only use MB when I can’t do otherwise since some of their templates are not easily re-created on Vegas.

>don’t shoot flat : shoot it right on the set.

Ah, yes, but that requires that you know what you want beforehand. Personally, I don’t decide the kind of color grading that I want (or even the overall look and feel of the video) beforehand. I only decide when I have the footage shot, because when I arrive at the place where I shoot, I don’t know what to expect. For the music videos I shot so far, I’ve never seen the location before! I had to improvise on the spot, after arriving there. It’s the curse of doing things the cheap way, I guess. Therefore, shooting flat is the right way for me, because it gives me the flexibility to go one way or another in post.

>DVD & 720p web delivery

It looks good on my TV as I said. I never export at 1080p though, just 720p so far. It looks good when rescaled down to 720p. I don’t expect to get great full 1080p from the 5D anyway, since the real resolution of the camera in video is at around 1600 pixels, and not full 1920px. So why bother with 1080p exports… 90% of the population can’t say the difference between 720p and 1080p from 2 meters away! Besides, I hate the 1080p rendering times on that PC.

And let’s not forget that my previous camera, the HV20, had a real resolution of 540p after being interpolated when I was exporting out a 60i video (very few times I shot in PF24 with the HV20). So going from 540p, to 720p (downscaled from the actual 960p that 5D/7D do), is already a major step forward for me! Again, everything is relative in this world! :-D

>such a big influence on wannabe filmmakers

The key word here is “wannabe”. Most people who read this blog are serious but amateur videographers. Their shooting style is similar to mine. So I simply tell them what I do, and they’re free to follow the advices, or not. I’ve done well for myself so far in my shoots. I have no regrets over decisions I made, or complaints of quality and such. I’ll certainly won’t advise my readers to rent a RED though. We’re not the demographic. ;)


rhervag wrote on January 25th, 2010 at 3:16 AM PST:

i was saying on a 52 *feet* screen at the theatres, where most readers want their hardwork to land someday somehow weither released or at festivals…
as digital projectors get installed everywhere, it is pretty cheap to rent such a commodity for a single night private screening if there is good advertising to gather enough local interest.

for the rest, i agree that having to improvise is the opposite way to do it for me especially being on no budget schemes. most of the time i’d rather sit and discuss concept and issues beforehand to build up the idea, know what to expect and aim for the best. looks get approved upon reviewing similar materials and a stylesheet helps envision the project and act as a guideline for all parties involved. being a one mand band relying on casual helpers doing it on their freetime out of passion needs strict rules to enforce a zero budget approach but still reach a common goal.
btw, never had the budget to rent a RED too :) but as those get more popular it will be cheaper soon enough


Daveson wrote on January 25th, 2010 at 9:35 AM PST:

C’mon, we don’t want to compare a 5d to the RED.
The 5d is killer for the money, and i’ve seen music videos on mtv that were shot with a 5d. so what. it can always get better.

cheers.


William Eggington wrote on January 25th, 2010 at 11:23 AM PST:

Are RED cameras the only ones that can do higher color depth full frame video capture?


David wrote on January 25th, 2010 at 12:25 PM PST:

The custom ‘flat’ styles definitely give you a lot more play in post production, much easier to expose in low light too. Though the problem I have found shooting with a custom ‘flat’ style is that the tones between your mids and your shadows (lets call them mid-shadows) tend to get noisy and often muddy, especially in low light. Introducing this much noise in camera really isn’t much better than starting with a more contrasty though cleaner image and tweaking it a heap in post. Dynamic range means very little if 1/4 to 1/3 of your entire image turns to digital grain. I’m still undecided whether its better to shoot with one of these ‘flat’ styles or to shoot ‘Neutral’ with sharpness, contrast and saturation turned down. No doubt though, that both options are better than shooting ‘standard’ setting.


This is the admin speaking...
Eugenia wrote on January 25th, 2010 at 1:28 PM PST:

>tend to get noisy and often muddy, especially in low light.

This is true. But think of the alternative. I have. :D

The alternative is to take the low-light video, and change the contrast in post, in order to get back shadow detail. You know what you get in that case? Not sensor noise, but h.264 artifacts! And between h.264 artifacts and sensor noise, I prefer noise.


Chris wrote on January 26th, 2010 at 11:11 AM PST:

>tend to get noisy and often muddy, especially in low light.

Absolutely! I’ve been filming three movies on my 7D in flat along with my XDCAM and was so disappointed. Just before dumping it I decided to give it a lost shot in “neutral picture style”. This converted to Cineform which gives you 4:2:2 out of the h264 is the best way to shoot HDSLR. Marvels Cine mud and flat mud is not what I want. As far as color grading goes this is clear: do primary on set, secondary in color.


rhervag wrote on January 26th, 2010 at 11:52 AM PST:

i understand that some enjoy to tweak things in post to no end and a few even take great pride themselves in being colorists but after all the internet debate is done and workflows are tested in the realworld, there seems to be an emerging opinion that wanting to replicate what Hollywood does with perfect colorspaces from capture to projection on much more expensive setups is not where those DSLRs really shine.

of course it’s fine for DVD and web ouput where the downsampling, compression and smaller resolution will smooth out most of those problems anyways. but there are sufficient issues like moiré, aliasing and jello to deal with on set that get worsen by such heavy grading workflows up to ruining usability for proper projection and distribution after the footage was thought to be secured.

That’s the only reason why I just wanted to throw a word of caution about that to balance the obvious enthusiasm for such method as I’ve already had to reconsider them times and times again when ‘near theatrical’ quality results were expected and results not up to snuff. i’m not being critical about the idea as i wished it worked, just weigthing against it in the current state of affairs. I mean i have mojo and MBL like any other aspiring shooter, but please Canon give us more bits or 4:4:4 and we’re talking again…

for the time being, securing a correct exposure via ‘neutral’ with dialed down settings and postwork kept to an absolute minimum proves ALWAYS to give much more pristine looking imagery when projected at full resolution on a big screen for an audience than having to shift around a compromised dataset with lesser precision.

even some of the people who created and released those famous flat picture profiles in the 1st place have now reverted to using a tamed neutral on most occasions after bad surprises and headaches when back from the battlefield. what if the extensive gradation ruins a shot with more artefacts and re-shoot is not an option ? getting it right on location at least saves the day for all parties involved.
the ‘fix that in post’ mentality is not an universal solution and calling for trouble in the first place.

so while i’ll agree for the most part, beware of the gotchas to be able to deal with unexpected results


Ryan wrote on January 27th, 2010 at 9:53 AM PST:

Emily,

This is great. My complaint with the superflat style was exactly that — the pink faces it produced. I’m going to add “ExtraFlat” to my just-published manual, “The DSLR Cinematography Guide,” here:

http://nofilmschoool.com/dslr


Ryan wrote on January 27th, 2010 at 9:56 AM PST:

Oops. That’s ‘nofilmschool’ with two ‘o’s, not three. Try that again:

The DSLR Cinematography Guide


Burk wrote on January 28th, 2010 at 12:17 AM PST:

Howdy Eugenia, great site. Really fascinated by all the color grading discussion. I’ve been shooting a lot with the 7D and like many I’m trying out different settings and tweaks. Right now I’m working on a zero budget zombie thing and am playing with the color grading.

I would love to hear your opinion on this:

http://www.vimeo.com/9032620


Mikeymike wrote on January 28th, 2010 at 5:48 AM PST:

Just wanted to post that you can achive the teal look with two vegas plugins:
color corrector:
Low: Angle=286.5 Magnitude=0.437
Mid: Angle=305.5 Magnitude=0.133
High: Angle=128.7 Magnitude=0.397

Cookie Cutter:
Circle
Feather=1, Border=0, RepeatX=1, RepeatY=1, Size=0


Mikeymike wrote on January 28th, 2010 at 5:51 AM PST:

Sorry forgot some of the color corrector settings:
color corrector:
Low: Angle=286.5 Magnitude=0.437
Mid: Angle=305.5 Magnitude=0.133
High: Angle=128.7 Magnitude=0.397
Saturation=0.944
Gain=1.102
Gamma=1.314
Offset=-14.2

Cookie Cutter:
Circle
Feather=1, Border=0, RepeatX=1, RepeatY=1, Size=0


This is the admin speaking...
Eugenia wrote on January 28th, 2010 at 2:01 PM PST:

Sorry, but using the cookie cutter around faces is the wrong way of doing this. Especially since Vegas has no auto-tracking abilities like After Effects has.


Nick (mrphoto) wrote on January 29th, 2010 at 4:55 AM PST:

Interesting experiments with your picture style. One thought – try basing this on the Canon ‘faithful style’, as opposed to ‘neutral’.

Both have very similar characteristics except that faithful is colorimetrically accurate at 5200k – accuracy of blues is a definite improvement over the neutral style IMO


Ben wrote on January 29th, 2010 at 9:07 AM PST:

rhervag is absolutely right here. If you’re projecting your footage in large format, get the look as close as possible in-camera before you record. You can get away with minimal color grading after the fact, especially if you’re playing around in Cineform’s 4:2:2 colorspace. But the 4:2:0 just doesn’t have enough chroma information to grade completely from flat.


MikeyMike wrote on January 29th, 2010 at 1:02 PM PST:

Eugina, the cookie cutter is just used as a darkening tool there is no hole to lighten unless you want (I could have used contrast control instead). Try it. I guarentee you will see the same result.


Daveson wrote on January 29th, 2010 at 1:57 PM PST:

Sorry Eugenia for being off topic, but i coudln’t find a simple (cause english is not my mother tongue) answer.
I bought a 7d and Neoscene on PC, and after converting the mov files they are much brighter (more noise)!
Can you help out?
Thanks!
and thx in general for this great blog!!


This is the admin speaking...
Eugenia wrote on January 29th, 2010 at 2:44 PM PST:

Daveson, it’s normal. Cineform uses a different gamma. Use a bit punchier picture style in that case.


Daveson wrote on January 29th, 2010 at 2:53 PM PST:

Wow, that’s bad.
So when i record, on the display the image seems to be too punchy, but than it’s ok after conversion! Not that nice, is it? My pic style is setup as Tyler Ginter discribes on vimeo.
Contrast and Sharpness all the way down and Sat 2 ticks left.
And the Original Mov looks good to me!

THX


Chris wrote on January 30th, 2010 at 6:08 AM PST:

Here is a quick with and without Cineform:
http://blog.mxr.at/cineform-for-7d/


Daveson wrote on January 30th, 2010 at 3:31 PM PST:

Thanks for the link, Chris.
Think I’ll have to learn more about gamma, brightness, noise etc.
still the original mov looks better to me :-(


Chris wrote on February 1st, 2010 at 12:45 PM PST:

@Daveson: Of course the original h264 \looks better\. But it doesn’t give you any headroom to process the color. For postproduction you’d want something like the cineform output.


Daveson wrote on February 2nd, 2010 at 3:11 AM PST:

Thanks Chris for the info.
Got my 7d just a week ago, now it’s time to learn!
Bookmarked your blog!


Comments are closed as this blog post is now archived.

Lines, paragraphs break automatically. HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

The URI to TrackBack this blog entry is this. And here is the RSS 2.0 for comments on this post.