[Marc Wielage]

I think the answers to your questions boil down to: you need to examine your philosophy of color correction and how you approach the problem, and alter that to fit the software. Don’t expect that the software will alter itself to what you need. This is also true for Baselight, Lustre, Mistika, Pablo, and everything else out there, not just Resolve.

I find for tracking, sometimes I can get a fairly decent track completely automatically; sometimes it only works for part of the clip; and sometimes it requires total frame-by-frame roto. It is what is is. I try to make the right decision for aspects like softness and shape and get them right the first time, so I don’t have to worry about having to change it mid-way through the shot.

Sometimes, multiple shapes work better than a single shape, particularly when you’re working on a tight deadline. I have a couple of pre-made shapes that I use that will fit a lot of shots (for example, head and shoulders), and this seems to help.

[Marc Wielage]

[Chris Wright] “o my.. you’re hosed. prores is not DPX log, EXR half float, or Tiff 16bit. you need at least 16 bit capture or your throwing away dynamic range below 16 f-stops. maybe its 8mm? lol.”
For a lot of old film, I don’t think you’re right. I think the film itself is hosed. And I’ve worked on literally hundreds of film restoration projects. Given lots of time and money, I’d absolutely choose 16-bit DPX, but it’s not always possible from a time and money point of view. And there’s dozens of ways to color-correct it, not all of them using curves. A lot of film made before 1980 is not going to have nearly the dynamic range you might think it has on paper. Real-world, it’s just not there — it’s all noise and grunge.

The real problem is dealing with the damage, unsteadiness, flicker, uneven grain, and density shifts within the frame. For that, you pretty much need some kind of pixel-by-pixel mapping within each color channel. And that’s pretty specialized software. We had a ton of flicker and density shifts on the 1977 Star Wars (when I was on the restoration team at ILM in 2004), and it took a team of about 20 people to solve those problems. The color correction was the easiest part; it was the everything else that was the problem. That was a DPX job, unfortunately only 2K, but I expect the film’s new owners will do a new 4K scan and do a much better job.

[Marc Wielage]

I have no answer for the crashes — so much depends on hardware configurations and source formats — but if I need to do something like this, I set up three parallel nodes with keys set to grab specific luminance bands. Top one for highs, middle one for mids, bottom for blacks — and then adjust those accordingly. You could also use a YUV splitter node to sharpen just luma only.

The new ResolveFX Sharpen Edges and Soften & Sharpen plug-ins also allow some very tweaky kinds of sharpening that are very good at excluding everything you don’t want to grab. I tend to use these in combination with tracking masks for certain soft photography issues.

I specifically don’t use LAB because it slows me down too much and I’m generally on a pretty tight schedule. I could see it being used to create a prefab look for a very, very specific situation, but beyond that, it’s too “fiddley” for me.

[Marc Wielage]

[charles wing] “Does it make ANY sense to transcode it to a higher res for color correction in Resolve? Or employ any other workflow modifications to get the best results? “
Not to me, no. It’s just a very simple 8-bit YUV format. You could transcode to ProRes 422 (or an equivalent DNxHD format), but I wouldn’t necessarily believe you’re going to get anything more out of it. You can’t get more out of 8 gallons of water by pouring it in a 10-gallon barrel.

There is a new de-banding tool in Resolve 14 that can selectively eliminate the ugly 8-bit artifacts that sometimes happen with these formats, and it does help somewhat. I find a lot of noise reduction and sometimes frequency-selective blurring & sharpening can also take the edge off.

[Marc Wielage]

[Joseph Owens] “Always good to evaluate — going in, the model introduced yesterday seems four-walled to the Apple application ecosystem, ie FCX, Compressor, Motion.”
Naaaa, I priced a “reasonable” (i.e., no 4TB SSD boot drive) version and it was around $9K, which I think is about right. There’s an article on the web I can’t find at the moment that disassembled an iMac Pro and then priced all the components, and said it would cost about 85% as much to build this same machine as a PC. So there is an “Apple Tax,” but not as much as you might think.

[Marc Wielage]

[Tero Ahlfors] “Life’s too short for trying to control things you have no say over so I grade everything to a specification and when it’s out of my hands then that’s that.”
These are very wise words.

[Marc Wielage]

I think the “Duke & Tero Show” could be either a new sitcom on the CW, or a new cartoon series on Adult Swim.

[Marc Wielage]

[Alan Okey] “By virtue of its design, the iMac Pro will be thermally constrained, and I know that I certainly wouldn’t want to be one of the first guinea pigs who gets to discover whether or not the new model is up to the task.”
The lack of upgradability is also a big concern. I think there’s a big market for the iMac Pro for certain kinds of users, like offline editing, sound editing, and DIT work, where an all-in-one system has its advantages. I don’t think it will be enough for a really demanding Resolve user. I’d also like to see what’s going to happen when the Resolve GUI hits a 5K screen like this.

Having said that, I just got a second Trashcan Mac as a stopgap just to get me through the next7-8 months’ worth of projects. I’m in a “wait and see” mode as far as the new Mac Pro goes, and am hoping it gives post people better alternatives than the iMac Pro.

[Marc Wielage]

Mixed framerate project are a huge red flag to me.

My suggestion would be to consider an alternative workflow: for example, transcode all the clips to a common timebase, so (no matter what the original speed) they’re now ready to go for a specific deliverable format. For example, if you’re going to deliver 29.97fps for broadcast, go with that. Be warned that Optical Flow yields different results on different systems, so it may be a frame earlier/frame late depending on circumstances.

If you have a single 422HQ file, that should not be choppy assuming fast storage. I used a relatively modest RAID0 for some time that I think only had about 330MBps, and I had no problem doing 2K 24p ProRes 444 projects all the time. Depending on the node tree structure, I could use 4K within a 2K timeline, then switch to 4K for delivery.

I think it also helps cutting the material up into smaller bite-size pieces. A continuous 2-hour or even 90-minute file is hard for some drives to handle. I would cut it up into (say) 4 x 23-minute files and see how that does. I would also split it up into at least 4 timelines if possible and see if that also improves performance.

You might also upgrade to OSX 10.12.6 and see how that does. I’ve had pretty good luck it and a Trashcan Mac over the last few months.

[Marc Wielage]

[Duke Sweden] “So which is it? Stabilize or not? Anyway a lot of the advice here helped. I like where it was pointed out that tracking is never perfect “except in tutorials”. That was what was killing me. Now that I know these things I don’t mind manually tweaking bad tracking. btw, I’m 63. I know what life is like. I’m used to it.”
I dunno, I think the tracking is actually a little better for me in 14 than it was in 12.5. One key is sometimes it helps to track a slightly-largish area, then shrink the window down for the final effect.

I just did 6 or 7 shots of eyeball tracking of a creature in a horror movie I’m working on, and it was about 90% perfect. It still required some manual tweaks, but it was mostly right on the first time. One key thing to understand is that the tracker works on contrast, so if you can pick a point with a lot of contrast, that will help. Trying to track dark, muddy shots is tough.

Often, what I do is qualify a key, then use a window to “contain” the key and protect it from outside details. Other times, I’m using a window to just give the actors a little bit of additional fill light (and/or sharpening). Manual tracking doesn’t scare me at all — it’s just a thing. I figure if the automatic tracking even takes 50% of the work out of it, that’s something.