Filmier Than Film: How Poker Face Cinematographer Steve Yedlin, a Self-Taught Coder, Makes Digital Video Look Like Celluloid
March 24, 2023

Written by Lane Brown, Vulture

A lot of recent TV shows share a similar look: flat, dark, and desaturated. As a half-informed aesthetic snob, I had assumed this was due to the limitations of the digital cameras that these shows were shot with — but it turns out that only partially explains the problem, because then I saw Poker Face.

Rian Johnson’s Peacock series, which stars Natasha Lyonne as an amateur detective who can identify murderers by their micro-expressions, was also shot digitally, and yet it has the highly pleasing and once-unmistakable appearance of traditional film. The colors are warm; there’s grain, distortion, halation, and gate weave; images sometimes drift artfully out of focus. Johnson says Poker Face was inspired by classic TV procedurals like Columbo and The Rockford Files, as well as the paranoid movie thrillers of the ’70s, and in some scenes it could pass for one.

Much of the credit for Poker Face’s visual template goes to Johnson’s longtime cinematographer Steve Yedlin, who shot the first and ninth episodes. (They’ve also collaborated on Looper, Star Wars: The Last Jedi, and the Knives Out movies.) Yedlin is a self-taught coder who has developed his own algorithms that can make digital video look like real film — enough so that it’s impossible for most people to tell the difference. (Over on his website, Yedlin shares an impressive demo of his techniques.) On a recent Zoom call, Yedlin discussed Poker Face, convincing Johnson to go digital, and how to spot bad cinematography in a car’s brake lights.

If I were to summarize the Steve Yedlin philosophy of cinematography, it might go something like this: It’s not so important which camera you shoot with, because any modern professional camera, digital or film, will collect the same basic visual information. What really matters is how you transform that data into your final image. Unfortunately, most cinematographers don’t do much transforming, so they’re at the mercy of their camera’s factory settings. But you’ve invented your own process by which you can imbue digital video with the qualities we like about film, including deeper colors and grain. Is that right?
That’s approximately right. I would add this: Some people assume that a filmed image just inherently looks a certain way, and when I do my thing, I’m altering or faking it. But I’m not. It always has to go through some process, no matter what. Otherwise, it would be like if you shot a film negative and never developed it. There’s always a transformation from the pure digital sensor or film negative data into the rendered photograph. I’m just actually using that leverage point.

So digital cameras come with their own built-in math for transforming raw image data, and you replace that built-in math with math of your own, which responds in a more nuanced, filmlike way?
This is a little technical, but most people could probably understand it if they wanted to. In digital cameras, most of the preloaded color transformations are using something called a 3x3 matrix. That sounds fancy: “Matrix?” It sounds like a mesh that you put over the color volume and it knows exactly what to do in every corner of the volume — in this part of the color volume, blues get desaturated, but in that part of the color volume, blues get more saturated. But actually, it’s the opposite. It’s the simplest, bluntest thing you could possibly do. So the first thing I do is undo the manufacturer’s prebaked matrix stuff.

What does a matrix-based color transformation look like?
With a matrix, everything is either going out or everything is going in. So if subtly saturated colors like skin tones, which are sometimes close to gray, get twice as saturated, then a car’s brake light, a neon sign, or a green laser will also get twice as saturated, which is too much. I’m sure you’ve seen a movie or a TV show with a car brake light, and it’s like, “Whoa, that looks really electronic.” Part of the reason is matrix math.

Is this all done in postproduction? Or can you see what it looks like in the monitors as you’re shooting?
The color transform can be applied as an LUT [a lookup table, or a digital file that transforms an image] so we can see it in the monitors on set. You don’t really need to see the spatial stuff — grain, weave, halation — until later, so we’ll do some of that on the dailies and then the real deal on the final. People ask me, “Why do you do all this stuff? There’s all this effort.” But when I’m shooting, I’m actually doing less of this stuff than anybody else. This stuff is all figured out beforehand, and once it’s set up, it just goes through.

How many other cinematographers do you know that write their own code?
Not a lot. But while I’m obsessed with this stuff, and I have a whole website dedicated to it, obviously the main thing that determines what something is going to look like is the actual shooting of it. This image rendering is the last little polish on it.

Let’s talk about shooting Poker Face. It’s been compared to the work of Gordon Willis, who of course did The Godfather and All the President’s Men. I’ve also heard Rian Johnson say he wanted it to feel like a throwback, and that he used Columbo as a reference. How did you settle on its look, and how did you pull it off?
It’s nice of people to say it looks like Gordon Willis — I wish it was Gordon Willis — because I was definitely doing a Gordon Willis top light in certain scenes in Adrien Brody’s office and in the crow’s nest, the room that overlooks the casino. In the crow’s nest, Rian said, “I want it to be dark in there.” So I did a thing where it was lit from far away and asked him, “Do you like this?” And he was like, “Yeah, whatever.” We hadn’t talked about it in advance. We don’t discuss the details too much. I just know that he likes things to look dramatic.

Rian and I have been doing this together for so long that we don’t spend a lot of time having abstract wine-tasting conversations about references or how to make something feel like it’s from the ’70s. It’s more about just rolling up our sleeves and designing the particular shots. For example, in episode nine, “Escape From Shit Mountain,” there’s a scene where Joseph Gordon-Levitt is dragging Natasha Lyonne on the road, and he’s bathed in the red light from a car’s taillights. Rian had that idea, and so I had to figure out how to practically do it — which red to use so it looked good and not electronically off-putting.

A big part of Poker Face’s look comes from the emulated film grain. Were you using more than usual? Is Poker Face grainer than, say, Glass Onion?
It’s actually the same as Glass Onion — kind of. Glass Onion was the only thing we’ve ever done where there were two different grain levels, one for theatrical and another for streaming. I hope to not do that again. Poker Face has the same grain as the streaming version of Glass Onion, and perceptually less than the theatrical version.

Why did you make two different renders for Glass Onion? Was it because of the way Netflix’s compression algorithm processes the grain?
I don’t want to go too far down this road, but yeah, it was. The grain was making the image look crummy. You could see really big compression blocks and swimmy stuff. There are actually different settings that would’ve made the grain in the theatrical version of Glass Onion look fine, but Netflix can’t change their settings just to make one show or movie look better. They have a meat grinder that everything goes through.

There are certain directors who believe that film will always look better than digital and can’t be convinced otherwise. But you changed Rian Johnson’s mind. He’s said he wanted to shoot The Last Jedi on film, and then you sent him a 30-page dissertation arguing for digital.
I did, and then he said, “Thanks for doing this, but no, we’re shooting on film.” Rian was very much in that film camp and said absolutely no digital. But by the end of Last Jedi, we had transitioned from mostly film to mostly digital. So that movie is about 50/50. Knives Out was the first all-digital one, where he flipped from “Okay, I’m acquiescing to this” to “I’m really excited about it.” One of my big selling points to Rian was the ability to stay in focus. In the film days, every single movie had blown focus. It was just a question of how much and how bad. But with digital, it’s possible to do harder shots with shallower depth of field and have no blown focus at all.

What’s kind of amazing is that Poker Face looks more filmic than some TV shows that are actually shot on film. For example, Succession looks great, but its colors are pretty denuded, so I had always figured it was shot digitally. But it turns out they shoot on film.
I don’t want to speak to Succession because I don’t know their process. But sometimes people believe that the mere act of shooting on film magically gives something a certain look, and so they don’t do anything else to control the pipeline, and then they end up using the same tricks that they would with a digital camera. And on top of that, there’s just a Zeitgeisty way of how a lot of things look right now. Sometimes when people say they want something to look “filmy,” they’ll see it and decide they don’t want it after all, so they’ll try to make it look like other things.

A lot has changed about the way film is made and developed. Does new film stock look as good as the older film that you’re emulating?
Film now isn’t what it was before. There used to be a huge infrastructure for it. Now it’s this little side-fetish boutique thing, but it’s not something that can be run in a boutique way. You need people who know how it works, and those people are leaving the labs.

People think I’m like the digital guru, but I’m really the most film-crazy person there is. Most film zealots aren’t doing the work — Kodak made the film, and they just use it. But I’m so into what Kodak and Fuji did over 100 years that I’m doing everything I can to continue in that tradition.