Panasonic Gh4

Getting A Film Look – Visual Workflow

 

It may seem counter intuitive to start babbling about workflows, but I think if you have a decent workflow it will allow you to consistently get a semi decent look from your footage. Keep in mind I shoot on a Canon, so all my discussions will be centered around my use of Canon.

Canon Picture Profiles

There are so many different picture profiles available for Canon DSLRs that you can easily get lost. By now I am sure you have heard about the Technicolor CineStyle picture profile. It was designed to use a LUT and mimic the old Technicolor look, but most people use it without applying the LUT (including me). Personally, I feel the obsession over CineStyle is unwarranted.

The Problems With Technicolor’s CineStyle Picture Profile

For 8 bit video (which is what most DSLR shooters have to work with) you get 256 values per cRGB channel (which are numerically assigned from 0-255, with 0 being black). Now, for us video shooters this sucks. It sucks because we want infinity color values per channel! With more color you get less banding and a truer image. Anyway, Technicolor decided to clip the blacks and so the CineStyle picture profile only sees color from 16-255. What this means is that by using CineStyle you technically lose detail (usually noisy, ugly muddy detail) in the darks which you will never get back. The truth is it is very rare to get pixels with values below 10. So visually, if you were to crush everything below 10 or 12, you would not notice the difference because your pixels never go below 10 anyway. So technically you get a lower dynamic range but its not a big deal. So when the creator of Marvel’s Cine Style picture profile starts crying about how Technicolor’s CineStyle sucks because of the loss in dynamic range, don’t pay much attention to it. Oh wait, he already did that. Clipping the blacks at 16 is fairly aggressive, and you can spot the difference if you do direct comparisons. But I would not lose much sleep over this.

No One Picture Style is Perfect

It is impossible for one picture style to be good at everything. Literally. Lets say you have a picture profile that has 10 stops of dynamic range. The luminance values in 8 bit video range from 0-255. Different picture styles allocate theses luminance values differently to each stop of light. If a picture style wanted to distribute the luminance values evenly to each of its 10 stops of light, that would mean each stop would have about 25.5 luma values to work with. This would be a pretty crappy strategy because the underexposed stops are very noisy making it difficult to spot detail. So trying to pack a lot of information into the highlights and darks results in wasted information. This is a limitation of the physics behind imaging sensors, as well as the camera’s software. Therefore it makes sense to capture more luma information in the stops of light that are in the middle. Each picture profile out there has a different take on allocation info within its dynamic range. So when you use a picture profile that has a wider dynamic range and has more detail in the darks, it comes at a cost. Ever wonder why some picture styles make skin tones and faces look like garbage? It’s because the picture style is allocating information in a way that reduces detail in the range critical to skin tones. Technicolor’s CineStyle is a culprit of nasty skin tones. Furthermore you will often encounter banding issues when working with Technicolor’s CineStyle.

How Much Dynamic Range Do You Need?

If your scene only has 7 stops of light, and you are using a picture profile that is designed to capture 12 stops of light, you are wasting a ton of information. Your footage will look like crap because a ton of the detail that has been allocated to the 5 unused stops is not being used by you. So your video is still technically 8 bit, but in reality it is only 7 bit, or 6 bit, or 5 bit! You won’t be using the full 0-255 color depth! It could be that you are only using values between 50-150 with the high dynamic range picture style. If you used a less aggressive picture style (maybe one optimized for 9.5 or 10 stops of dynamic range) you would capture more information; maybe you would be capturing a range of values between 25-225 or something. Basically it comes down to this; because DSLRs shoot video only in 8 bit, you only have 256 tonal values for a gradient. The higher dynamic range of the picture style, the farther apart those gradient tones get spread. To get the most out of the 256 tones, use a picture style who’s dynamic range matches the dynamic range of your scene.

To truly see what each picture style is doing you need to conduct some tests and look at the histograms, waveforms, and vectorscopes. Now if you are like me, and are to lazy to figure anything out for yourself then you should google the picture profiles that you are interested and see what type of IRE numbers show up for them in the waveforms (IRE stands for Institute of Radio Engineers). Basically IRE are the units by which waveform monitors measure pixels of light. The white parts (highlights) are located around IRE 100 and darkness is around IRE 0. Canon cameras can actually go above IRE 100, but the internal codec clips the values at about 100. In my opinion, stuff below IRE 30 is not too important. Neither is stuff above 85 IRE. That is not to say that you should ignore the data in those ranges. I am just pointing out that for my color grading needs I am more interested in the useful parts of dynamic range (IRE 30 – IRE 85).

You Only Need 3 or 4 Picture Styles (max.)

  • Midtone Priority: This would ideally be your default picture profile. This profile would be used when you favor color over dynamic range (for me that’s almost always). This type of profile will favor midtones over shadows and highlights. Furthermore if you were to look at a waveform of such a picture style, you would see that between IRE 25 and IRE 80 (the vicinity where useful dynamic range lies on DSLRs), the profile would fit fewer stops of dynamic range than a high dynamic range picture style would. As I mentioned before, you are trading dynamic range for higher quality data. I often see a boneheaded mistake on YouTube. People hosting web shows, and using tons of cheap lighting. They light their faces, and use a low contrast/high dynamic range profile to record their faces. Why in the hell do they do this???? They don’t need the dynamic range. They are shooting their face with a three point lighting set up for crying out loud! They need accurate skin tones. Usually, they don’t grade the footage very well and everything goes to shit.
  • Skin Tone Priority: Skin tone priority will behave similarly to color priority. If you really need to get the skin tone correct, I would just go ahead and use a default portrait profile.
  • Dynamic Range Priority: If you really care about dynamic range, than this should be your default profile. It should be a somewhat balanced profile that can handle midtones. The whole point of having several picture profiles is so that you always have the right tool for the job. Ideally one of your high dynamic range profiles will perform well enough in the midtones so that you can use the profile as a default.
  • Even More Dynamic Range Priority: Never use this as a default profile! If you truly need low quality dynamic range for whatever reason, use this picture style. You can even turn on Highlight Tone Priority, a mode which gives you an extra bit of latitude in the highlights, to go along with this profile to get even more dynamic range (at the price of added noise).

As you can see, you can easily get by with two picture styles:

  1. A style with good color rendition and low banding artifacts
  2. A style with a bit more dynamic range.

And if you don’t care much for dynamic range, well then it’s even easier for you. Just shoot a nice balanced style.

Recommended Picture Styles:

  • Neutral: Shane Hurlbut, who was the DP for Act of Valor (most of which was shot on Canon 5d mkIIs) used Canon’s Neutral picture style for all the 5D shots in that film. From what I understand, he likes to lower the contrast as well as the sharpness when using this profile.  Its a great all around profile.
  • Prolost Flat: This is basically the same as number one. Stu Maschwitz, who is a DSLR blogger and cinematographer/editor (has worked on too many films to name), uses a setting that he has dubbed as Prolost Flat. It uses the Neutral picture style with contrast and sharpness dialed down. Phillip Bloom also uses Neutral as his base picture style.
  • Flaat_10p: This is one of the picture styles offered by a very cool person named Samuel Hurtado. The 10 stands for ~ 10 stops of dynamic range, and the p stands for a Portrait. Samuel Hurtado has multiple picture styles that he has developed, so check out his website for more details. Flaat_10p is optimized for skin tones. So if you are really concerned about getting good skin tones, use Flaat_10p, or the Canon Portrait picture style. If you want less dynamic range and better colors, go for the Flaat_9 series.
  • Technicolor CineStyle: if you like to have fun color grading footage, this is a fun picture style. As I mentioned above, CineStyle doesn’t have any special features that make it “the best” picture style. If you understand the limitations behind 8 bit video, you will understand that every picture profile has strengths and weaknesses. If you are trying to get great color tones, options 1-3 are better profiles for the job.
  • Flaat_12: If you need as much dynamic range as you can get, use something like flaat_12. I only recommend this for extreme circumstances. Despite having a number 12 in the name, it only gets about 11.5 stops of dynamic range.

I will some day go into further detail about picture styles because it can get pretty advanced. Maybe I will do a shootout and post histograms, waveforms, and color graded results. But for now, take my word for it; there is no magical picture style. Choose the right tool for the job. And if you are not sure about what you need then stick with what the pros use;  Neutral profile with low contrast and sharpness.

Editing:

Next thing you want to do is drag and drop your footage into After Effects. You want to set your project settings so that you are working in a 16bit or 32bit depth. The reason you do this is so that when you add effects and grade your footage, you are not limited to the 8 bit color values (0-255).

When you add your footage to the timeline, the first thing you should do is denoise the footage. The best consumer noise removal tool out there right now is a software called Neat Video. By removing digital noise, you will somewhat mitigate the ugly digital look of your footage, and also have the ability to edit your footage more aggressively. When you denoise your footage, you introduce banding. The reason for this is that areas with subtle gradients previously had variation in the color (noise), now that the variation has been removed, those subtle gradients are more solid/constant making banding more apparent. To remedy this problem we have several options:

  1. Add color grain
  2. Blur individual channels
  3. Use the ramp scatter effect
  4. A combination of all three

I recommend playing around with all three options and seeing what produces the best results. After you have cleaned up your footage, you are ready for FilmConvert.

FilmConvert takes your DSLR footage, and transforms it to look as if the footage was shot on film stock. This is achieved by applying grain, and messing with the colors. You can use it as a stand alone software or as a plug in. The color adjustments is simply to make it look as if you shot the footage strait to a film stock and not to a digital camera and is not intended to be a substitute for color grading.

By now your footage should be free of ugly digital artifacts, have the properties of whatever film stock you chose to mimic, and ready to be color graded in a 16 or 32bit color depth.

Color Grade Your Footage

If you plan on grading in Davinci Resolve (which is free and really cool), render out your After Effects project in one of three formats:

  • DPX
  • TIFF
  • Cineform
  • ProRes
  • Uncompressed native (h.264 .MOV for Canon)

By rendering your video in DPX or TIFF, you avoid degrading your video quality via compression. Exporting DPX files is a more advanced option. If you don’t know what you are doing, you can easily screw up your files when rendering them out. I would not recommend using DPX unless you already know what it is and how to use it. TIFF files will be lossless and, like DPX files, result in extremely large file sizes. If you don’t have the hardware to handle it, opt for Cineform or ProRes. Cineform and ProRes are the two most popular intermediate codecs for amateur filmmakers. If you plan to stay in After Effects, then just leave the footage alone, proceed to the next step with your FilmConvert adjusted footage.

If you want to get even fancier, let Adobe Media Encoder handle all the encoding. In general, working in lossless is a good idea if you are doing professional/paid/commercial work. If you are uploading to Vimeo or Youtube just for non commercial reasons, it is a waste of time and resources. You would be better off working with lower quality files, and spending the saved time and headaches on shooting more footage. My recommendation is this: use cineform, or just work with the original files until the final render.

The Film Look Color:

Today everything is greenish/bluish. I think if you are going for a big budget, crime, thriller, action, or political thriller, than this look is fine. But don’t over do it on the blue/green. If you saw my film look examples or my Criterion look examples, you will see that this is not the look we are going for in most cases. The everything is blue except for skin fad is something that has only become popular in recent years. If you look at older films, you will see that they don’t have this crazy color grading. Here is what the current trend looks like:

378834-4

hunger-games-movie-katniss-pretty1

movies_rihanna_battleship_still_5

frame 01021605R

safe-house06

safehouse-movie-picture-31

avm-5-1

the-bourne-legacy-movie-still-27

skyfall-movie-still-45

It looks like a bunch of barbie dolls were placed into some environment where everything is blue and green. My brain hates it. For some reason, the creators of these films do not want us to see many colors. They only want us to see blue, green gold, yellow, brown, etc… The look is muted, lacking any sort of richness. If you are making a bleak film, or something that is focused on action then this is not a big deal, I guess. But if you are going for something that looks closer to a genuine film look (not, modern digital action genre look), stay away from this aggressive look.

Big budget studios are not the only culprits of making films that look… blah. Indie filmmakers are worse offenders. They often go over the top with the colors, but usually its the ultra low contrast that irks me. Have you seen the ever so popular low contrast look on Vimeo? It mostly sucks. Just surf Vimeo long enough and you will understand what I am talking about.

Since this is just a workflow guide, I won’t say much about color grading. I guess my theory is simple; make your colors balanced. If you overdo or underdo the color, saturation, and contrast it can look unappealing. You can always use reference footage or photographs as a reference. There are multiples tools that allow you to color match footage (Davinci, Color Finesse, Colorista, etc… So if you have a movie that has a look you strive to emulate, you can import it into whatever program you are using, and color match your footage to it.

Adding Film Grain:

After you have finished color grading your footage, you should add a little bit of extra film grain. The reason I prefer to add even more grain (remember, we dithered the footage with grain, then FilmConvert added additional grain) is because I upload my videos to the internet. And if you read my page about YouTube and Vimeo compression, you will see that when you upload your file, it gets heavily compressed. I often notice that the grain I added to my footage does not show up very well on the internet, so by adding more grain than usual, I am anticipating the effects of compression. I recommended using the film grain provided by Rgrain or Garilla Grain for this job. Both Rgrain and Gorilla Grain are cheap solutions for getting good looking grain. You get multiple grain styles packaged as video loops. By using video loops, you make rendering and editing much less stressful on your computer hardware. If you want film grain that is real and doesn’t cost an arm and a leg, buy Grain35 by CrumplePop. If you want grain that is even better than Grain35, go with CineGrain or IndieScans. CineGrain and IndieScans provide robust packages of very high quality film grains that are very easy to edit. Unfortunately their prices are too steep for amateur filmmakers working with tiny budgets.

If you don’t have much money and are in a scenario where you have to choose between 8mm, 16mm, or 35mm grain, go for 35mm grain. 35mm grain is finer, less obtrusive, and is the grain you see in most films.

This page is a work in process, and hopefully I will add on to it. I hope you learned a little bit about making your digital footage look more film-like. The workflow I outlined is very common and there are multiple resources on the internet that go into further detail about editing at a high bit depth. The last piece of advise I have is: figure out how to compress your file for YouTube and Vimeo so that all your hard work doesn’t go to shit!

  3 Responses to “Getting A Film Look – Visual Workflow”

  1. Hi there!

    About what was said concerning switching to 16-bit/32-bit workflow in AE to work with more than 8bit color values, when I do that my footage turns faded?

    I understand most dslr’s shoot in 8-bit (which I have), is it possible to convert this footage to be 16-bit so when I grade in AE I have more colour values?

    Thanks!

 Leave a Reply

Login With a Social Account

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Film Brute