Marc Wielage

Sorry for the late reply: I had an unexpected day off and just saw this question today.

The standards for film scanning are kind of nebulous, and I have seen vast differences in levels from film scans done by Technicolor, Fotokem, EFilm, ILM, Cinesite, and other companies; even NY and LA divisions of the same companies can be different. I just eyeball it, come up with a reasonable setting for Offset (Printer Lights), then add a node for a Custom Curve (generally an S-shape, but it depends), then subsequent nodes for balance and level, sometimes a few RGB tweaks if necessary. Once the picture is in kind of a "faux Rec709 zone," I color correct it as if it were digital. The same expectations and values still hold: it's not that much different from digital, really.

It does help if the filmmakers shoot color charts daily, and you can pull those up and see how far off or how close you are to reasonable hue and vector values. I was always skeptical about the Macbeth color chart in a digital world, but the DSC charts are good, and the XRite Color Checker Video is not bad (but I'm suspicious of the saturation levels). You can get a lot just out of a grayscale chart, which will at least help you balance out the blacks and whites.

I've remastered 32 films shot-on-film this year, and sometimes I have to throw my approach out the window because either the elements, the scans, or the actual exposure is just completely whacked. I find I use Printer Lights more than I do with an all-digital show, but other than that, it's essentially the same thing. Modern film emulsions are just about as clean and sharp a 4K digital, done right. 

Eh, my usual deal is punch the saturation, pop the whites, consider a soft roll off for highlights, a vignette, and a little film grain. That's a typical "commercial look" to me. Nothing wrong with that, anything to aggressively grab the viewer and make them watch what you're trying to sell. 

I had some issues with my original dongle from 10 years ago, and finally hit on the solution of plugging the USB dongle into a 1-foot USB extension cable. After that, the cable itself took all the abuse while the dongle stayed fairly pristine. I needed this when I had to use other client machines or had to move sessions around sometimes.

It helps to have multiple dongles as backups: between Resolve and Fusion, I think we have at least 7 or 8 copies at the office, and I have a spare in my laptop bag (including a Linux Advanced dongle).

Anybody who's been at this for more than a few years has had to deal with long shots or long single-shot projects. I did a single-shot music video back in 1984 (Steve Perry's "Foolish Heart"), and I used what we would now call keyframes and about 20 color corrections for it.

In more modern times, we have access to more complex methods. Steven J. Scott did an interview five years ago discussing what he did on Birdman, and though this was all done on Lustre, the principles apply to any color correction software. It's a great-looking film and Steven did an amazing job on it.

Gamut Mapping is a wonderful feature. I use that shot-to-shot if I run into scenes with intense car tail lights or (say) a scene in a nightclub with neon signs. It does something similar to a hue-vs-sat curve, but it targets it in a more subtle way. I don't like the idea of legalizing a show with a "one size fits all" clipper, but an occasional Gamut Mapping node will do the job when you see an illegal gamut excursion on a Gamut scope or a Diamond scope. 

I have had whole scenes of 70 or 80 shots that was all nightclub interiors with intense background signs, and for that, I'll either do a group grade with a Gamut Mapping node or I'll have a fixed node tree with one node just dedicated to legalizing. That's assuming I'm delivering Rec709. If we had an HDR pass, I'd dupe the session and create a new version with no Gamut Mapping and just let it blow.

I sometimes combine the dissolve transition with a color dissolve (or even several color dissolves) just to change the levels as needed. This is very hard if you have a flattened file -- as with a film scan -- but is relatively easy to do with two separate files. Basically, you're dealing with a non-linear dissolve, and you can move around the keyframes as required. 

2 hours ago, Rakesh Malik said:

Nowadays I use power windows a LOT less than I used to, because I learned to sculpt and control light a lot more on set because having basically screwed myself, I knew what to look for the next time around.

Bear in mind that Richardson was working on a $100 million dollar movie, so he had the time and budget to have his crew go in and fix a lot of lighting issues on set. In the real world, I think power windows and masks and keys and all that are still very necessary. To me, Tarantino's approach is unnecessarily dogmatic (ala Dogma 95), and the trick to me is make all the VFX and color work completely invisible. If you can't tell there was ever a window there, then it works. The moment you can see evidence of it, then the illusion is shattered. 

On the other hand: I think Once Upon a Time in Hollywood is a brilliant film, it looks great, it's very entertaining, and I've paid to see it twice. I can't offer any higher praise of a movie than that. Tarantino has earned the right to have strong opinions on where and how he does things, and that extends to post.

Jesus, if it's absolute silence, it's deadly. On the other hand, there are those people who get distracted by music. Jan De Bont did not want to hear anything in the background when I did one of the home video versions of Die Hard in the 1990s. He was not a cheerful man -- but his cinematography was spectacular. Ultimately, he liked the work, and that's all I care about.

21 minutes ago, Anton Meleshkevich said:

I never listen to music while grading. Either original sound of a project or some podcasts.

Oh my god, I'd be dead if I couldn't listen to music while color-correcting. I've been doing that since the very beginning (provided we don't have sync sound). I always let the client choose the music when possible. Apple Music, Spotify, and Amazon Music are godsends -- and if the client thinks it's too distracting, we have "the sound of silence." But it actually goes faster with some quiet background tunes.

16 minutes ago, Anton Meleshkevich said:

@Marc Wielage Actually I'd really like to know how deep you (or other colorists) go into problem solving with total rotoscoping and relighting.

It depends! There's a terrific podcast interview with the great Bob Richardson on Once Upon a Time in Hollywood that came out last week, and he spends 10 minutes talking about Yvan Lucas' excellent color-correction. One thing that Tarantino insisted on was: no power windows. I can't imagine doing a movie without them nowadays, but Richardson and Lucas clearly figured out a way to solve the problem, and quite a bit of what you see was just "lit that way" right on the set. It's brilliant work.

Podcast link:

https://www.hollywoodreporter.com/behind-screen/once-a-time-hollywood-dp-featured-thrs-behind-screen-1226416

I mainly use windows to knock down the background and add shadows and depth to the image, but everybody works differently. When you're under a tight schedule, you're not going to have time to digitally relight every shot (ala Steve Scott's excellent work on The Revenant). Even at my end of the post sewer, I still will use a window here and there to pop up a face or flag off a harsh background, just because I can. I sleep better at nights that way. When there's absolutely no time and money, we wince and let it go. There are certainly limits to how far you can push things, even with raw camera files and/or original negative. We stand on the DP's shoulders, and unless they do their job, we can't really do ours.

22 minutes ago, Anton Meleshkevich said:

This LUT is called "Great DP and art direction" 😎

No, the CO3 colorists are world-class people who do impeccable work. They pretty much set the gold standard for the business. (As do the Steve Scott's who use Lustre, Maxine Gervas and Yvan Lucas on Baselight, and so on.) These people contribute greatly to the ultimate look of the picture, and they also have the ability to work calmly with clients and solve problems, which is an exceptionally difficult skill to cultivate over time. LUTs are at or near the bottom of the list. Great lighting, great lenses, and great exposure all help considerably.

You'd have to ask Mark. The link for the page with a shot of his node tree is here:

https://www.hurlbutacademy.com/director-of-photography-why-do-we-color/

My guess is that it's a key on the window. One trick of these fixed node trees is that you have to be very careful of anything adjusting level, because that will affect keys on down the signal path. For that reason, you either have to break out Parallel nodes or just structure the keys earlier in the chain. There are valid reasons to go either way.

I have had cases where I pull a soft highlight key, qualify it carefully, and then will boost the levels a bit just to "pop" the scene more. I just had this happen a few weeks ago on a day-for-night scene where I had to drastically darken the shot, but wanted the headlights alone to pop out at a normal (for night) level. This had to be tracked carefully, but there's several ways to do it. Highlights are helpful sometimes, as are the Log controls -- it kind of depends if you're coming up or coming down.

I have generally duplicated the shot on the edit page and then cut in a shape with an alpha to cover up the flaw, then color correct on that second clip to match levels, and then soften the shape to make it blend in better. Patch Replacer essentially does the same thing automatically, but without the benefit of manual control. Both are useful under the right conditions.

This is often a bone of contention in the film restoration business. My take is you need to do NR last, because the contrast that happens in correction could exaggerate noise problems more in some cases than others. If you apply the NR as the initial node and then correct after that, it's better for caching but you will see unequal noise levels caused by different settings cut-to-cut and scene-to-scene. I think it's a decision that has to be made differently per project.

I'm generally a fan of not noise-reducing unless we really need it, so I'll do it scene-wide for X number of shots, but then turn it off once the exposure goes back to normal. Added grain is kind of a separate issue.

I haven't seen this problem in 15.3.1 on Mac OSX 10.14.5, either with the Mini Panel or the Advanced Panels.

How many nodes? What specific hardware are you using? I typically use anywhere between 15 and 30 nodes, but there's a lot of "it depends" in there. I do run with quite a few nodes bypassed when they're not needed, but I certainly do a lot of enable/bypass actions when I'm doing a trim pass.

Another thing you could do is qualify a key on upper-mids and highlights, soften it to reduce artifacts, add a little NR, and desaturate there. The effect is different than a Lum vs. Sat curve and (to me) not as destructive if you're very careful. I do this all the time when I need to subtly go in the opposite direction and desaturate blacks without artifacts. But I try not to push it too hard.

Noted DP Steve Yedlin has some things to say about the "film look" and digital cameras, and he has some interesting theories and conclusions:

http://www.yedlin.net/OnColorScience/

I would point to the Oscar-winning film The Artist as an example where they not only digitally created "halation," but they went a step further and made modern 35mm color film look like 1920s B&W nitrate film. That's a very clever trick, and I think it helped sell the look and period of the film very well. Only Richard Deusy as Duboircolor in France knows exactly how they did it, but I think you can bet there were carefully-qualified keys, a glow filter of some kind, and some blur here and there. Great lighting and filtration in-camera probably helped as well. It's a great-looking film -- all shot on film, but (ironically) done at Red Studios on Cahuenga Blvd. in Hollywood. 

BTW, here's an interesting interview I missed a few months ago where Stefan was part of Panasonic's introduction of their new GZ2000 OLED display:

This is the longest I've ever heard Mr. Sonnenfeld speak about anything. 

Stefan doesn't give many interviews, but he did chat with Steve Hullfish's book The Art & Technique of Digital Color Correction, and I found a lot of what he said enlightening. My take is that Sonnenfeld works very quickly and uses simpler techniques than a lot of people suspect. He does extraordinarily good work, and there are few other people in Hollywood (and everywhere else) who grasp the importance of client relations, sales, business, and technology and have the ability to balance all of them the way Stefan Sonnenfeld does. 

Yes, you can grab a still frame of the "good" image and then paste that within the keyframe area and that should recall the keyframed color correction. You have to kind of figure out a strategy for dealing with color fading. To tell you the truth, I don't think Resolve (or Baselight or Lustre or whatever) is the best tool for that job: you'd be better off doing kind of a "best light" and then sending the files through an MTI DRS or a similar restoration system just to "stabilize" the color fading, and then take those files and color correct them for a final look. 

The problem with film color fading is that it's non-linear, meaning that one side of the image is going to be more faded than the other, so you could wind up having to use a lot of power windows and masks on top of the keyframe issue. The labor involved would be ridiculous, to the point where it'd drive you mad. 

One thing I know that Lowry Digital did back in the day (when I was there around 2010-2012) was they would break the image down into RGB, then process and reduce the flicker and fading on a channel by channel basis. Once that was done, they would merge the RGB files back together again. That may be more complicated than what you want or need to do, but it's interesting to note that generally one channel is flickering or wavering more than the others, and that might give you a clue as to how to attack the problem.

Yes, Dan Moran is very good. (He's also of normal height, which I like.)

I don't often get into "beauty grading," but most of what I know came from Dan's lessons, and they helped me immensely on a couple of projects. Dan also had some great ideas on how to mask and track complex objects, which boiled down to multiple shapes -- and that was a good lesson to learn. MixingLight is a terrific resource.

Naw, we go with just the 2-up Display at the moment. It's been fine. It's rare I ever go beyond 25 nodes. 18 is more normal for the way I work. There are always exceptions: a few weeks ago, I did a 90-minute documentary in 12 hours in 5 nodes, whole thing, maybe 900 shots. It's more a time/budget exercise, not "how many nodes can I make?"

The wide displays give me a headache because I'm breaking my neck on the keyframe window. I also like breaking out the external scopes to a 3rd display.

You could also paste the clip on a dedicated timeline on the edit page and just repeat it over and over and over, with a fixed Composite Level and setting. There are pros and cons to either approach. Note you can also color-correct the grain to intensify it or reduce it.

Walter is a very bright guy. I also use a fixed node tree and (because great minds think alike) I also have preset Left / Center / Right masks when needed plus a vignette.

One thing I got out of this was I've always had a TRIM node at the end for client trimming, but at Walter's level, he needs a DP Trim mode, an Exec Trim mode, and a Director Trim mode, which I had never thought of. That's a clever idea. 

Walter has said many times (so this is no secret) that it's very important to get a solid "balanced" image early in the image chain, and that's all part of the Node Order of Operations. As an alternate approach, look at the similarity of this technique to Mark Todd Osborne, who is also a very fine colorist here in LA:

I do a variation of this with some parallel nodes early on right after normalizing the image with a node for Offset/Printer Lights and a second node for a Custom Curve and/or individual YRGB pots (with Lum Mix at 0), followed by 16 more preset nodes. Most of them are turned off 90% of the time and I only activate them when they're actually used. 

But then there's days when this happens...

(That would be a joke. I do not actually use this many nodes.)