After recently completing the work on the second spec Walk MS PSA, “For the Children” I thought a discussion on grading shots in uncontrolled environments was in order.  As a filmmaker, you always want to plan, art direct, and control your compositions, but when you are running and gunning during a live event like Walk MS, you don’t always have the opportunity to stop the action so you can light it or art direct it to your “vision.”

Case in point, the following shot from the PSA featuring a child on his dad’s shoulders. Click on any of the pictures to see it in full 720P resolution.

Not a bad composition, but if I had the time and the crew, I would have loved to have put a reflector in there to brighten up the kid’s and his dad’s face.  I’d also like to get rid of that hand in the bottom left that’s taking the direct sun and becoming a focal point.  While I’m at it, I might have put a sky filter on to enhance the sky which over-exposed a bit.

So, should we let the fact that we didn’t have that time and crew in the moment ruin this shot?  Absolutely not.  Instead we  begin the  post color grading process.  Some people (including myself) call this color correction, but I think that grading is more appropriate as we’re doing a lot more than simply correcting the color.  For those wondering I shot on a Canon EOS 7D using Stu Maschwitz’s picture profile settings from my previous testing.   This flattens the image out giving you more highlight and shadow detail to work with when grading, which came in very helpful here.

I should also mention that I transcoded the original 7D quicktime file to Cineform files.  As I mentioned in my previous post the native Quicktimes are very compressed and have limited color information.  Cineform does a great job at preserving the original image characteristics while converting to a 10 bit 4:2:2 color space which makes it more robust for color grading.

I graded in Adobe After Effects, with Magic Bullet Looks because it works in 32 bit color space and allows me a lot of control.  You can also color correct in your NLE, or dedicated color correction program.   I will be making reference to tools in After Effects, but the principles will be the same no matter what tool you use. When you start to grade a shot you should remember to do things in the order that you would have done them if you shot for the look in camera.  First compose, then light, then add filtration, then apply the look or color timing.  The DV Rebel’s Guide has a whole section dedicated to this and why, I highly recommend you check it out.

So let’s start by doing what I wish I had done on site,  and add a virtual reflector to  brighten up the kids face.  I used Magic Bullet Looks (MBL from here on out)  Spot Exposure tool for this, but you can also use After Effects built in Exposure effect and feathered mask.  Mask  his face with a soft  feather and increase the exposure until it looks correct to you and the kid takes on the focal point of the shot.  Here is my result.

Not bad, but now Dad seems out of place.  Let’s bring his exposure up as well, without taking primary focus off the kid.

Better.  Now let’s deal with that bright hand drawing our attention off screen left.  I approached this the same as their faces, only using a virtual flag to decrease exposure on the hand.  While I’m at it, I also take the exposure down on the kids arm.

Much better. Now that I’ve re-lit the scene, let’s think about that sky.  Since this is a PSA that’s hoping to make you feel good we want a nice clear blue sky instead of the grayish overexposed sky we have here.  Again I  used MBL’s gradient filter to add a blue gradient over the top right corner of the sky.  This makes the sky blue without affecting the rest of the image.  You can also use a blue solid in After Effects with a feathered mask and transfer modes to create similar effect.

Now that I’m happy with the relighting and filtration it’s time to color time the shot to give it  it’s overall look.  As this is a PSA where we wanted to instill hope, I wanted to go for a warm look, so I gave the shot a warm, golden color cast and added a curves effect with a film-like “s” shape for contrast.

This gives it a warm, magic hour feeling.  Also notice how the look altered the blue of the sky, but in a way that’s natural to the rest of the shot.  If I added the blue after the look, it wouldn’t have taken on the golden cast, and would feel disconnected from the rest of the world the shot took place in.

Lastly, I shoot with a picture profile in the 7D that turns the sharpness all the way down.  I do this because out of the box the 7D (and most DSLRs and video cameras) over sharpen the images they shoot.  I don’t like giving my camera that much control. I like to turn that sharpening off or in this case down as much as possible so I can precisely dial in the right amount of sharpening for every shot.  For this particular shot I used After Effects “unsharp mask” filter.

But wait a minute, that looks terrible!  All the artifacts from the original heavily compressed quicktime are accentuated!   (If you can’t tell, click on the image to see the full rez version.) That’s what I said the first time I applied some post sharpening. While Cineform does an excellent job dealing with the color space, it doesn’t do any noise reduction or deartifacting. But don’t be disheartened. Simply go back to your source file and add some noise reduction as the first effect in the chain, or the first layer if you’re using adjustment layers. I used the noise reduction filter that came with Magic Bullet Steady and it did an excellent job. Other choices are Neat Video, or DE:Noise.  If you don’t want to buy a plug-in you can also try After Effects built in Reduce Grain or Median filters, but I couldn’t get the footage as clean as with the plugin, your mileage may vary.

Ah, that’s much better.  And to see the results in a much more dramatic comparison, watch the grading progression video below.

I have held firmly to the belief that when shooting digital video, it is prudent to shoot with a picture profile or camera settings that gives you a flat clean digital negative allowing for room to create the final  look in post.

This was only reinforced after reading the excellent DV Rebel’s Guide by Stu Maschwitz.   Some people disagree with Stu and I and think you need to lock your look in camera as this will not only save you post production time, but when you bake in the color it’s embedded in the camera before it’s compressed to file (or tape), meaning less potential noise in the image.

That’s all fine and good, until you’re editing, and you realize you need the emotional push of a scene to change or it needs a little boost.  Now if you want it to be cool rather than warm, or even worse less contrasty, you may not have fluidity to make that change, and your efforts to have less noise may now be yielding more noise as you push and pull the colors and contrast levels.

This is one of the reason’s why people still love shooting film, or using the RAW files from the Red One, or in their DSLR stills.  By working with the RAW sensor data, these formats give you the opportunity to change your mind, and process the image to your liking much like traditional film negatives.

In my previous camera the Canon XL2, I researched forums like DVinfo.net,  and was able to cull most of the users preferences for creating a flat look.  I tried them all and shot tests and mixed up different variables and ended up with my recipe of parameters to get my flat digital negative, and it served me well. That said, it wasn’t always rosy, there were moments during post on Mercy where I wish we had baked in a little more red to that “digital negative”,  to help make the skirt and mask pop as much as Tony wanted.  I had to push it pretty hard in post, and that resulted in some noise and artifacting, one of the drawbacks and concessions of shooting digitally to a compressed 8-bit medium like DV.  That also taught me the lesson to shot flat, but with a look in mind, and to always shoot tests to achieve that look before shooting as you may find you want to tweak your flat profile to achieve it.

So when it came time to set up my new  Canon EOS 7D,  I wanted to create a new baseline digital negative profile.  However I was concerned, because it records to an even more highly compressed, 8-bit, 4:2:0 color space format, and I didn’t want my color corrections falling apart. It didn’t take long to learn that I wasn’t the only one with concerns, or trying to find that flat digital negative.  In fact, I found that there are lots of custom profiles out there for creating a flat look, including one from Stu himself. (His settings were actually for the 5D but they work just as well on the 7D.)  I also found just as many arguments against this due to the codec in the camera breaking down and causing noise and color artifacts in the footage, as I had already feared.   As with everything in life, your mileage may vary so as I did before with my XL2, I downloaded all the presets, read the forum posts, and then shot some tests using each of the profiles to see for myself.

For my tests  my Wife art directed some of her antique Kokeshi dolls on a tray with some fruit and vase in the background, giving me a nice interesting and colorful still life to shoot.  I setup the 7D on a tripod with the canon plastic fantastic f1.8 50mm lens. (I know I should upgrade the lens, but I already owned it, as it was just fine for stills. One step at a time, people!)  I shot indoors with a soft mid morning natural light coming into the room, and a little fill bounce.  Set the aperture to F1.8 for some bokeh.  Set the camera to the built in neutral setting for exposure. Set video to 1080/24p,  shutter to 1/50, and ISO to 640.

TIP: I set exposure by rolling 2 seconds and then playing it back to see the histogram, in camera.  Adjusting until I got it right.  Man, do I miss zebra stripes.

Above is the image, with and After Effects levels effect showing the histogram in 32-bit mode.  This is the original untranscoded file, and you can see the “gaps” in the histogram showing the lack of color information from the 4:2:0 sampling.  My exposure technique obviously still needs work.  I’ve had the camera a week so I’ll cut myself some slack. I didn’t clip the blacks, and the whites are right there on the hairy edge with a little highlight clipping.   This profile does what it says and gives you a neutral image, but with Canon’s built-in contrasty gamma curve.

I used a laptop tethered to the camera so I could upload the many picture styles quickly.  Then without changing the aperture, iso, or shutter I cycled through each of the following camera settings and rolled 5 seconds of video for each one.

The standard profile obviously is much more contrasty and saturated.  You can see I lost some latitude in the exposure as the highlights clip. (The tall thin line on the right).

Stu’s settings, the only one that uses the in camera controls only.:

It  looks pretty good.  It achieved our goal of added latitude by pulling in the exposure to optimum levels, and flattened out the contrast without squashing the midtone details.   I tried a few aggressive Magic Bullet Looks (hereafter referred to as MBL) presets, and it held up well (see the bottom of the post for an example).

The Genesis Panalog WIP 4, is definitely still a work in progress.  It doesn’t look like a panalog file to me, it crushed some of the blacks, but rolled off the highlights OK, and seemed to over saturate the color. When I attempted the same MBL presets as I tried on Stu’s settings it did not hold up well, showing artifacts and banding in the saturated colors.

The Marvel Cine 1.2 picture style, was designed to emulate the Cine-Gamma curve of Sony EX1. Much like Stu’s settings it flattened the contrast without losing midtone details but didn’t really gain as much in latitude based on the levels results.  It held up OK, using the MBL presets, and  I obviously didn’t shoot with real skin tones for this test, but I did notice in some others that skin tones would get orange.

Superflat was designed by a Neil Stubbings on Cinem5D, and it gave me similar results to Marvel Cine, but it featured more strange color shift, and the same color artifacting issues that Panalog suffered.  In fact Neil has since posted that he’s stopped using his own profile and now uses Sharpness all the way down, contrast, all the way down, saturation down 2 clicks.  Exactly Stu’s settings above.

ExtraFlat by Eugenia is a modification of the above Superflat, and really seemed to be a winner giving apparent additional latitude, and an even flatter less contrasty look than Stu’s settings, without the color shift.   It also seems to hold up under aggressive color correction from the MBL  presets.  However, I really had to play with the correction a lot more, and push the effects alot further to achieve the same results with ExtraFlat as with Stu’s settings. Additionally I saw that ExtraFlat doesn’t pull the sharpness down all the way which I would modify if I were to use it.  I find it’s better to dial that sharpness in later in a more controlled post environment.

Overall I am leaning towards Stu’s settings because they held up so well and the ExtraFlat had to be pushed much further than I’d like which could create more potential for noise.

I’ve also found that I can achieve even better results by transcoding the native camera .mov files to Cineform files.  When transcoding to Cineform, the codec conversion interpolates the color space to a 4:2:2 10-bit file, that’s not only more robust to color correct, but easier to edit.

Since I started this test, some more very smart and experienced DSLR shooters have done some additional testing. Most significantly, the folks over at Zacuto did the The Great Camera Shootout comparing several DSLR’s versus 35mm film.  The results may surprise you, but what I found interesting was this quote from Phillip Bloom as to how they set up the cameras

For those who are interested I discussed at length with Tim Smith of Canon about the best picture profiles for the cameras. We settled on in camera settings, many will argue that we should have gone “superflat”, and I am sure it would have been great but in the end what we went for was sharpness all the way down, Contrast all the way down and Saturation down two notches. The reason we turn sharpness all the way down is simply the camera is too sharp and it also accentuates the flaws of the camera. It’s very easy to bring back a little bit of sharpness in post. That is my recommendation. With the flat PP I was able to get about 2 stops more latitude in these cameras.

From Philip Bloom’s Blog

Wait a minute… those are Stu’s settings…  I’m starting to see a trend here.

Then Shane Hurlbut blogged his feelings on color correcting DSLR footage, adding the concept of adjusting the white balance of your camera to remove any inherent bias, then using a very slightly modified custom gamma curve to create a RAW file.

I’ll have to give that a try next.   For now I’ll stick with Stu’s settings.   Below is an example of the shot with the MBL Blockbuster Preset, showing how it holds up to an aggressive look.  Mouse over to compare it to the straight camera look.

Happy shooting.

Scott Kirsner has posted an audio file of The Digital Revolution: What to Expect from Film’s Second Century on his excellent blog CinemaTech.

The audio file (which is unfortunately in RealAudio format) is about an hour and features Richard Chuang, co-founder of PDI/DreamWorks, Jeff Fino, co-founder of Wild Brain, John Knoll, visual effects supervisor at ILM, and Stu Maschwitz, co-founder of The Orphanage.

A must listen for the Digital as well as the analog storyteller, even if you do have to download RealPlayer.