Ask Mike Most Anything

[M Most]

How often do you use sharpening on Red or 35mm footage? Do you think this is a frequent or advisable practice?

I tend to stay away from sharpening in post unless specifically requested or absolutely required. It tends to produce a lot of artifacting if not handled in very small doses. In the film world, some sharpening is usually done in the scanner, particularly for 2K (and lower) scans. But in a grading environment, I've usually found that nearly all sharpening algorithms yield very "edgy" results, much like cheap video cameras. You can, of course, restrict it to specific areas of the image, such as the eyes, for instance. But while I agree that sometimes it's useful, for most properly shot images it often does more harm than good. Only my opinion, though.

[Marc Wielage]

But while I agree that sometimes it's useful, for most properly shot images it often does more harm than good. Only my opinion, though.

I would also add that I have seen many, many projects get kicked back from QC because of over-enhancement. It's way too easy for people to get carried away with excess sharpening. This can be a disaster for small features trying to deliver final masters to fulfill a distribution contract.

In my case, I've generally not had to use it more than (say) a dozen shots in a feature, but on those specific shots, it helped quite a bit. There are proprietary sharpening processes used by a couple of restoration companies I know of that don't apply enhancement in the usual way. When it's preceded by grain reduction, you can actually get away with more sharpening than you might think. The Bond movies on Blu-ray and the recent Star Wars video releases are good examples.

But this is not the same kind of tool as what's available in DaVinci, Baselight, or any conventional color-correction system. Enhancement can be a very, very dangerous tool in the wrong hands -- not unlike noise reduction. Too many people think if a little is good, then a lot is even better. Not true with enhancement and NR.

[Nathan Rosenquist]

Hi Mike,

Thank you for all of your contributions to this thread. We are lucky to have someone with your experience sharing their knowledge so freely.

I have a question about replicating the look of optical diffusion filters through digital means.

The conventional wisdom, as I understand it, seems to be that if you want to approximate the look of various optical diffusion filters you can get fairly close in post by copying your footage onto a new layer, applying something like a gaussian blur, and then lowering the opacity of the blurred image. In this way, you have both the sharp image underneath, and some of the blurred image blended in to smooth things out. Then both the blur and the opacity of the blurred level can be adjusted to taste to fine tune the effect.

Whenever I have tried this, it seems to work pretty well if all you want to do is soften the image up a little bit. However, I have found it much more difficult to replicate any of the other optical artifacts that some filters have (ProMist, etc.) In particular, it would be nice to be able to have selective control over halation in light sources. With the technique I described, there isn't really much halation at all. I also recall David Mullen saying in another thread that he shot a film (Big Sur) where they had only the blacks halate, drawing inspiration from older film techniques.

I'm especially interested in the general principles behind this, as opposed to an all in one plug-in or system that has a preset, as I'm trying to understand the differences at a more fundamental level.

Do you have any tips on how to gain better control over this process?

Thank you for your time!

[M Most]

I have a question about replicating the look of optical diffusion filters through digital means.

The conventional wisdom, as I understand it, seems to be that if you want to approximate the look of various optical diffusion filters you can get fairly close in post by copying your footage onto a new layer, applying something like a gaussian blur, and then lowering the opacity of the blurred image. In this way, you have both the sharp image underneath, and some of the blurred image blended in to smooth things out. Then both the blur and the opacity of the blurred level can be adjusted to taste to fine tune the effect.

Whenever I have tried this, it seems to work pretty well if all you want to do is soften the image up a little bit. However, I have found it much more difficult to replicate any of the other optical artifacts that some filters have (ProMist, etc.) In particular, it would be nice to be able to have selective control over halation in light sources. With the technique I described, there isn't really much halation at all. I also recall David Mullen saying in another thread that he shot a film (Big Sur) where they had only the blacks halate, drawing inspiration from older film techniques.

I'm especially interested in the general principles behind this, as opposed to an all in one plug-in or system that has a preset, as I'm trying to understand the differences at a more fundamental level.

Do you have any tips on how to gain better control over this process?

You can achieve specific effects by qualifying the area being affected by the 2nd layer. Luminance keys are the usual approach. By keying only the bright areas, halation effects can be had by varying the amount of the blur on the additional layer, and also by using different blending modes (add, screen, etc.). Conversely, the effect David was after can be achieved by inverting the key and affecting only the darker areas. These things can get considerably more involved, but basically that is usually the approach. Any one-button solution is going to have limitations. Doing it "the hard way" gives you a lot more control, so it's usually how professional colorists do it, both because of the superior results and because we're basically masochists.

[Marc Wielage]

The conventional wisdom, as I understand it, seems to be that if you want to approximate the look of various optical diffusion filters you can get fairly close in post by copying your footage onto a new layer, applying something like a gaussian blur, and then lowering the opacity of the blurred image. In this way, you have both the sharp image underneath, and some of the blurred image blended in to smooth things out. Then both the blur and the opacity of the blurred level can be adjusted to taste to fine tune the effect.

It's an ongoing debate as to whether to use filters during the shoot, or do it all in post. That goes for color, diffusion, grads, whatever. DPs often make a good point that doing it optically during photography is a more "organic" method, plus it bakes the look in to the point where nobody can take it out later.

My problem with that coming from post is that if they go too far with the filters in production, it's a huge problem to try to reduce it (or worse, take it out, if somebody change their minds). You have infinite control in post over how intense or subtle the effect is -- defocus, sharpening, tint, desaturating, etc. -- but that also means the DP has the potential of having to surrender control to somebody else later on.

Note also that using filters during photography can create problems with blue screen, 3D, and other VFX situations. The best answer I think is just to do a workflow test, try it both ways, and see what works best for you. If it were up to me, I'd almost always say, do it in post -- provided you have a decent budget and schedule for color-correction.

Mike above is absolutely right that pulling defocus keys has become standard procedure for a lot of stuff these days, especially with actors "of a certain age" and also controlling things like blown-out highlights and so on. A filter would treat everything more or less the same way; a defocus key could attack whites separately from everything else, skin tones separately from everything else, plus you could create garbage mattes (windows) to contain the effect only where you need it.

[M Most]

Note also that using filters during photography can create problems with blue screen, 3D, and other VFX situations. The best answer I think is just to do a workflow test, try it both ways, and see what works best for you. If it were up to me, I'd almost always say, do it in post -- provided you have a decent budget and schedule for color-correction.

Opinions are always based on one's own specific experiences, so perhaps surprisingly or perhaps not, I would probably say exactly the opposite, unless visual effects are involved - and yes, I consider stereoscopy a visual effect. But I'm basing that on the fact that I have had the good fortune of working with a lot of very experienced, very talented cameramen, most of whom started their careers before these kind of post techniques were commonplace, and thus have a very good understanding of what they're trying to achieve and how to achieve it. It's not necessarily the right answer for everybody.

[David Mullen ASC]

I've had less than satisfactory results trying to do digital diffusion for an entire show (as opposed to the occasional shot) -- I find it time-consuming and hard to keep the sharpness while getting that hint of halation around sources that I want. Whereas I know that if I use a 1/8 Black Frost, for example, that it won't impact sharpness much. With blur overlay techniques, it's hard to increase the halation while not blurring highlight details. I know it's possible but it so much simpler to just use the camera filter.

But my general principal is to under-do the level of optical diffusion and not use it at all on extreme long shots, or telephoto shots, or efx shots, etc. knowing that I can always add more diffusion in post. But for your run-of-the-mill dramatic scene where a mildly filtered look is desired, I don't see a problem with doing it in-camera as long as you don't over-do it. And the truth is that occasionally you get happy accidents with optical diffusion which is one of the reasons you use them, that sort of optical interactivity inside the glass, which feels very organic, just like how I prefer real lens flares to digitally-added ones.

[Michael Lindsay]

Hi Mike

How, witin the traditional telecine controls offered in systems like mystica/speedgrade (specifaclly those that do not have curves), is one suposed get fast and steep toe and shoulder compression that allows one to produce a really contrasty picture that keeps as much shadow and highligt details as possible?

In days gone bye I would ask for a particalur setup of the the leagliser (on the back of the TC suite) to take some of the super whites and blacks back in to the pictuture (heavily compressed). It was never particaully satisfactory but it would work a bit.

I am in a colour session on Mystica(s3d job) on monday and I am guessing I might have to resort to a fudged use of imported LUTs to provide what I want?

thanks for any thoughts you offer

Michael

PS I believe we once met in Miami. I was getting a tour of a facilty that was doing some processing for me and you where sitting at a scratch seat that was just newly put in...

[David Mullen ASC]

I'd also add that if you were shooting a TV series, with a ton of footage being generated, and the mandate was to make the female stars look as good as possible, saving any diffusion for the end of post wouldn't fly if you are going to be judged mainly by how it looks on the HD monitors on set and how it looks in dailies. By the time post is finished and the series goes to air, the DP may be long gone if he didn't make the female leads look good.

[M Most]

Hi Mike

How, witin the traditional telecine controls offered in systems like mystica/speedgrade (specifaclly those that do not have curves), is one suposed get fast and steep toe and shoulder compression that allows one to produce a really contrasty picture that keeps as much shadow and highligt details as possible?

In days gone bye I would ask for a particalur setup of the the leagliser (on the back of the TC suite) to take some of the super whites and blacks back in to the pictuture (heavily compressed). It was never particaully satisfactory but it would work a bit.

I am in a colour session on Mystica(s3d job) on monday and I am guessing I might have to resort to a fudged use of imported LUTs to provide what I want?

I don't have any real experience with Mistika, so I can't offer any specific suggestions on that. Some programs - Scratch comes to mind - have S-curve controls that offer a continuously variable S-curve, which would to a great degree accomplish what you're looking for. A lot of what you're talking about really comes down to having experience and touch. I've mentioned him before, but Stefan Sonnenfeld creates almost all of the looks he's famous for using not much more than lift, gain, and a few cleverly placed windows. It's not always about having specialty controls. It's often about using the basics to their utmost.