Protip: filmmakers, STOP “shooting flat” or using cinema color profiles on your camera

Update: I’ve written a much more comprehensive article on the subject of shooting with flat/log color profiles at the Gazing Cat Productions blog. This post is a good short example, but I show better examples and explain the problem in great depth at GCP. If you have the time and want to grow your knowledge about digital video, it’s definitely worth investing the time to read the 3000+ word post there instead of this one.

I plan to do a much deeper video on this subject later, but for now I’ve found a perfect example of the sort of “you should shoot flat [for that coveted film look of course]” bad advice I see online all the time. If your camera does not output files that have 10-bit (or higher) color channels, you should NEVER SHOOT FLAT, LOG, or CINEMA COLOR PROFILES, EVER. (Hint: if you don’t already know whether your camera produces 8-bit color or 10-bit color, it produces 8-bit color.) I’ll point you to this article  “Should I Shoot Flat and Underexposed?” with a Technicolor CineStyle example image so you can read their advice and then I’ll show you with THEIR OWN IMAGE why shooting flat on your DSLR is a really bad idea, then I’ll explain what’s really going on.

Here’s their original image which was shot with Technicolor CineStyle, a “flat” picture profile that supposedly helps you get more dynamic range and therefore better looking video:

Technicolor CineStyle example image
Technicolor CineStyle example image from The Association Blog

And here’s what it looks like after I pull it into an image editor, boosting contrast, saturation, brightness (a little) and gamma (a little) to make it look “normal” again:

CineStyle image after color correction
Color-corrected image. Note the blocky artifacts and unnatural colors.

Let’s take a closer look at the face side-by-side, before and after.

Original CineStyle, vs Corrected
Original CineStyle, vs Corrected, double size to show detail more clearly

Keep in mind that this is a small JPEG image from their website, not a lossless shot of the original frame. There are JPEG artifacts visible in both images, but those artifacts help us get a better picture of why “shooting flat” on cameras without 10-bit color is a bad idea: notice how the JPEG artifacts in the corrected image are WAY more obvious and the quality loss after my simple color correction drastically lowers the apparent “production value” of the image? That’s a big part of the problem, but the other part is the colors. The reduced saturation requires heavy saturation boosting to look normal again but the damage caused by discarding a lot of the color DIFFERENCE information cannot be undone; the face has color banding issues that make it look more “plastic.” The flatter you shoot, the nastier this color banding gets. No amount of correction or magical LUT will ever make it look normal again.

When you shoot flat, the 1/2-stop of dynamic range you gain comes at the cost of effectively ruining your color and increasing blocky compression artifacts and noise. That’s why you can’t color grade properly. That’s why it looks like garbage when you push it in post. It’s not your fault; you were given advice intended for $10,000 cameras and you’re holding a $500-$1000 camera. This applies to any flat, LOG, or “cinema” picture style. Technicolor CineStyle, Panasonic Cinelike-D and V, V-LOG, Canon S-LOG, all of them will permanently damage your 8-bit footage and possibly make it useless. You can’t outsmart the basic math: 8 bits of space can’t hold more than 8 bits of data.

Normal vs. Flat Profile
Stretching flat or log picture profiles back to normal curves results in banding artifacts. Once visual information is lost by shooting flat, it’s gone for good!

Use a standard picture color profile with all tweaks set to zero or their default values (turn noise reduction all the way down though, especially on Panasonic mirrorless cameras), take test shots, push the footage in your editor to see how far it goes before falling apart, tune your settings, and repeat until you get the best results possible straight out of your camera.

As I mentioned before, if you happen to have a camera that can output 10-bit color (to the files on your memory card, 10-bit HDMI without an external 10-bit or RAW-capable recorder doesn’t count) then you have 4x or more added color detail that will be lost in the 8-bit final product after editing anyway, but the only DSLR or camcorder I am aware of that is affordable to consumers and has 10-bit color is the Panasonic GH5. Even Canon’s expensive new 5D Mark IV DSLR only outputs 8-bit color! In the camcorder world, 10-bit color is available on the Panasonic DVX200 which is about 40% more expensive than a Canon 5D Mark IV.

If you want to read more, this excellent article does a great job of explaining further, including images that illustrate the problem of “breaking up the histogram” brought about by shooting “flat.” I will post a  video about this eventually, so subscribe to my YouTube channel if this topic interests you.

One final note: picture profiles only apply to compressed video formats. If you’re shooting in a RAW video format like many very high-end cameras can produce, you’re getting 100% of the sensor information already, so picture profiles simply don’t apply in the first place and you don’t have to care about any of this stuff for that camera. Of course, if you’re shooting on a $50,000 camera rig, you probably aren’t reading this post, either…

9 thoughts on “Protip: filmmakers, STOP “shooting flat” or using cinema color profiles on your camera

    1. Appeal to authority is a logical fallacy for a good reason. I don’t care what Philip Bloom thinks. I care if what he advocates produces good results.

      On any device that produces 8-bit depth output files, simple math says I am absolutely correct. If you squeeze a 32-level section of the brightness range into a 24-level section (no fractions) and then expand it back to 32 levels, you don’t get the original 32 levels back. There are extremely limited circumstances where log footage can achieve better results on 8-bit cameras, but that’s effectively limited to scenes of extremely high contrast where quality is less important than avoiding clipping.

      If you have 10-bit output files, you have a lot more detail than the final 8-bit output, so using tricks like log can work in more situations. Shooting “flat” (cranking down contrast and saturation and whatnot) is always a mistake.

  1. Hi, I’m wondering if it’s a good idea to bake in a color grade via picture profiles. I’m assuming DSLRs with 8 bit video apply the picture profile before the compression (hopefully), so in theory, would I be able to pre-color grade my footage and reduce the work in post production?

    If picture profiles are taken into account before the raw readout is compressed, would that mean it might be possible to create cinematic LUT-like filters and apply it to to get good colours/dynamic range without worrying about 8 bit issues like banding etc.? Obviously it results in loss of flexibility in post, but I’m just curious.

    1. It’s a neat idea, but I would advise against it in general. The problem is that you’ll be permanently applying that grade to the footage. If you change your mind at any point, you’ll be stretching the values out more than if you had just shot with standard/neutral color and graded from there. I don’t know if picture profiles are applied to the raw data before 8-bit quantization or not, though I imagine it is, but the “distance” between graded footage and differently graded footage can be significantly bigger than the “distance” between neutral footage and either grade.

      Keep in mind that “cinematic look” is largely a function of lighting, costumes, sets, locations, and camera handling. The film Tangerine was shot entirely on one iPhone. iPhones (in my opinion) are very inferior compared to even just a 10-year-old premium compact point-and-shoot like the Panasonic LX7. Despite this, they managed to create a “cinematic” film, because it’s all in the presentation, not the technical superiority. LUTs and picture profiles and 8-bit/10-bit are not the right path to go down until the stuff that’ll be in front of the camera is right. Get it right in-camera as much as possible without getting crazy, then worry about grading in post.

      Also, if you publish to YouTube or Netflix or any other online streaming platform, the video will have crappy compression artifacts anyway, no matter how pretty your 1080p master ends up looking. They don’t care. They compress to save bandwidth. You could tell them to set the x264 quantizer mode to AutoVAQ (Biased) [3] but they won’t listen.

  2. I’ve always wanted to make movies (short films, anyway) and only recently could afford to buy the 5-year-old Panasonic G7. I can’t say I have the resources to handle 4K or can competently grade even 1080p. Please, how do I get the best possible color straight out of the camera?

    1. As far as the picture profile goes: Standard profile, leave everything in it set to 0 except noise reduction which should be at -5. You don’t need to be competent at color grading; what you need to learn is to lock the white balance properly. DO NOT USE AUTOMATIC WHITE BALANCE. ALWAYS LOCK THE WHITE BALANCE. You can’t grade a constantly changing white balance without horror (I’ve done it). If you shoot with the wrong white balance, it’s easy to pull the footage into an editor and use something to change the color balance of the footage. Premiere Pro has the Lumetri Color effect which has tons of options, but just sliding the Temperature and Tint (blue/red vs. green/magenta) numbers will allow you to easily fix your white balance later. If color is not strong enough, you can boost Saturation. If color is too strong and you want a more muted look, lower Saturation. Color grading is an art and a science at a high level, but on a fundamental level, “color grading” is really just about selecting a look that feels right to you for the mood of the scene.

      Always remember that everything outside of the camera is more important than the camera itself. Sets, costumes, lighting, actors, good writing, decent camera operator. If you want to be that camera operator, practice shooting in fully manual mode with ISO, aperture (iris), and shutter (exposure time) locked. Practice manual focus and focus pulling. If you have an electronic focus lens (the ring isn’t mechanically linked to the focus inside the lens, which is the case if you spin the ring and it never stops) then you can’t do focus pulls and need to practice using the touchscreen and autofocus to change focus points in real time (called “focus racking”) and accept that some bad shots might happen due to the focus racking going wrong (repeat until it’s right, or don’t focus rack).

      I shot my first film on a Canon R60 consumer camcorder. It only has very crude “idiot level” manual controls and it has an absurdly small sensor. It is a capable camcorder but doesn’t produce any image that is remotely as nice as the G7 does. The camera isn’t important. I made several works with that camcorder before buying something bigger. Don’t worry about the camera; worry about getting good results.1

Leave a Reply

Your email address will not be published. Required fields are marked *