Ask Mike Most Anything

[jake blackstone]

Being lucky enough to work with two out three colorists mentioned, plus of coarse very brief time working with Stefen Sonnenfeld at POP, I would have to agree with Mike wholeheartedly, that nothing replaces the experience of learning from the best and putting in years of actual hands on training.

 

[M Most]

Being lucky enough to work with two out three colorists mentioned, plus of coarse very brief time working with Stefen Sonnenfeld at POP, I would have to agree with Mike wholeheartedly, that nothing replaces the experience of learning from the best and putting in years of actual hands on training.

Interesting that you would mention Stef. Under the category of "never stop learning," even though I've been at this for about 30 years now (my original mentor was Adam Adams way back in...... oh, probably before an awful lot of people here were born......), I had the occasion to do a pilot with Stef about 2 years ago. I spent about an hour with him and the director setting looks (Stef at the controls, of course). I found his approach and technique really unique and fascinating, and adopted a number of his methods for that and other projects since. So, as I said, there's no substitute for one on one interaction, and if you're smart, you never stop learning.

 

[Scott Crawley]

Thank you gentlemen both, for your replies.

 

[Marc Wielage]

Interesting that you would mention Stef. Under the category of "never stop learning," even though I've been at this for about 30 years now (my original mentor was Adam Adams way back in...... oh, probably before an awful lot of people here were born......)...

Ah, next time you see Adam, tell him hello from me. Adam will remember me, since I'm the guy who trained him (back at U.S. Video in 1980). :wink5:

 

[Peter Moretti]

Hi Mike,

I was wondering if you could riff a bit on two topics. One is using RedGamma vs RedLogFilm. I've heard it mentioned quite frequently that using RedLogFilm gives a more adjustable image even when referencing original .R3D files. That seems illogical to me, b/c both methods deal with the same underlying data, so I simply don't see how one gamma setting would offer more gradablily over the other. And it would seem that if one of the RedGammas gets the image close to where you want it to end up, then why not use that as a starting point?

The other question is about using curves instead of offset wheels (either in normal or log mode). Curves feel more intuitive to me, but it seems that they are not favored by many colorists. For example, last time I looked at it SpeedGrade it didn't even have curves. I'm sure the answer has to do with using the right tool for the specific task at hand. So may I ask, when are curves your weapon of choice and when are wheels?

Thanks so very much!!

 

[Corey_O]

Mike ,

What would a ball park budget be for a color suite with a theatre and Pablo installed?

Just curious to compare the old telecine rank suites with davinci installed.

I'm guessing newer should have smaller budget. Technology and all.

Thanks

 

[M Most]

Ah, next time you see Adam, tell him hello from me. Adam will remember me, since I'm the guy who trained him (back at U.S. Video in 1980). :wink5:

I lost track of Adam a long time ago. After he left Fotokem, I really don't know what happened with him.

 

[M Most]

Hi Mike,

I was wondering if you could riff a bit on two topics. One is using RedGamma vs RedLogFilm. I've heard it mentioned quite frequently that using RedLogFilm gives a more adjustable image even when referencing original .R3D files. That seems illogical to me, b/c both methods deal with the same underlying data, so I simply don't see how one gamma setting would offer more gradablily over the other. And it would seem that if one of the RedGammas gets the image close to where you want it to end up, then why not use that as a starting point?

The information in the orginal RAW version is the same, but the interpretation through the gamma curve is different. Since that interpretation is what is presented to you when you start grading, it does have an effect on both how you grade it and how much flexibility you have in doing so. If you use one of the Redgamma curves (any one of them, for the purposes of this discussion) you get a starting image that has been "normalized" into a video like gamma, as opposed to the log coded image that you get with RedlogFilm. Because of this, if your grading involves raising any lower end values, you will likely see more noise working from the Redgamma image than you would from the RedlogFilm approach. That's the nature of how grading works. Color correction is generally a destructive process, and concatenated transforms - particularly those that bring up lower end values - create issues that don't necessarily exist in the original image. Values in the lower end of the scale on a gamma encoded image tend to be noisy because there aren't a lot of levels available for quantization. When those values are transformed into higher ranges, that becomes obvious in the form of random noise. By starting with a log coded image, nearly all of the important values in the image are in a part of the curve where noise isn't a factor. Beyond that, there are the practical considerations of how one goes about grading. I have long advocated using log controls - namely exposure and contrast - as the primary controls for log coded images. In that scenario, you are not separating the image into different brightness ranges for the purposes of balance, you are instead altering the levels of the entire red, green, and blue records all at once. For that to really work as a grading approach, the image really needs to be in log space. So for me, I always prefer interpreting Red files using RedlogFilm because I feel you generally get a better result with log grading with respect to recreating the original photography and the original captured colors with better separation and fidelity. And that, for me, is a much better starting point for any creative interpretation that might or might not be necessary.

The other question is about using curves instead of offset wheels (either in normal or log mode). Curves feel more intuitive to me, but it seems that they are not favored by many colorists. For example, last time I looked at it SpeedGrade it didn't even have curves. I'm sure the answer has to do with using the right tool for the specific task at hand. So may I ask, when are curves your weapon of choice and when are wheels?

Like most professional colorists I know, I rarely use curves, for a number of reasons. Curves can be OK for an overall transform (i.e., converting log images to video or P3 gamma spaces), but unless you really understand what that curve is doing, and what it should be doing, it is far too easy to create artifacts unintentionally by using them. Things such as either "flat" or "too contrasty" faces are common when a curve is not precisely designed or controlled. For this reason, most colorists I know use a predesigned "precise" curve in the form of a LUT for this purpose. On the other hand, my friend Leo Marini (owner of Local Hero Post, and a really talented colorist as well) uses them all the time, primarily to make images "pop" and achieve a brighter, "snappier," and generally "happier" look. You can see this on some of the pictures he's worked on ("Pitch Perfect" is a good example), and for that kind of thing, curves can be useful, and I've sometimes used them for that reason. I will also sometimes use them for opening up the blacks slightly, or increasing midrange contrast. But those are all "touch up" scenarios. As a primary color tool, they are not generally used. For one thing, they cannot easily be controlled by a tactile physical control, like a trackball or a knob. This means that changing their values requires a mouse or pen approach, and for most professional colorists, this largely eliminates the "feel" that is so critical for tactile feedback and brings a lot of the talent of the colorist to the table. By the nature of their design, they also limit you to essentially changing only one thing at a time. For most colorists, grading is always an interactive process that requires the ability to change multiple parameters at the same time, both in terms of the primary color channels and the overall values. With a differential control, such as a trackball, you're never changing only one channel, you're always affecting all three. This, once again, allows you to "feel" your way to the delicate balance that you want. There are times when you might want to change only one channel and in most systems, that ability is also there. But for the most part, you're working in a world of balance, and that is where the "feel" part comes into play. With curves, this ability is lost, and you're constantly moving from one to the other, which is not only inefficient, but leads to inappropriate and misbalanced values. So the short answer to your questions are curves are never my primary control, but they are sometimes useful for specific things.

 

[M Most]

Mike ,

What would a ball park budget be for a color suite with a theatre and Pablo installed?

Just curious to compare the old telecine rank suites with davinci installed.

I'm guessing newer should have smaller budget. Technology and all.

Thanks

I don't feel it's my place to quote potential budget figures. But relatively speaking, a telecine room in the "old days" generally cost almost $1,000,000 for the telecine alone. A DaVinci 2K - the most popular color corrector of the HD era - was almost $400K. And a 32 inch Sony CRT monitor - the "Rolls Royce" of that era - was almost $40K new. You also needed clean room environments, an ultrasonic film cleaner with proper ventilation, much more electric power, elaborate signal routing, multiple different videotape formats, and a sizeable vault space - all unnecessary today. Today, the Pablo is among the most expensive - if not the most expensive - of all of the grading systems. So if you substitute the components needed for a DI theater, and compare that to an HD grading suite of the mid to late 90's (and most of them were not using Ranks by then, they were generally using Spirit telecines, but the cost was similar), you're exchanging a 2K for a Pablo, which is something of a wash. You're exchanging a $40K monitor for a projector that's roughly double that. And you're building a bigger room with a large screen. But nowhere are you still incurring the cost of a telecine machine, which was a million dollars all by itself. Nor do you require the costs of all the other items I mentioned. So the answer to your question is that it's far, far less costly today than it was then. But then again, the rates were considerably higher then, the deliverables were more costly (videotape machines were generally around $100K or more each, and tape stock was about $75 per hour), and in general, the business was at least somewhat more profitable, although post production has never been a particularly profitable business. Lower cost of materials and equipment does not always equal healthier or more profitable business models, not when the rates come down severely and not when your customers sometimes become your competitors.

 

[Corey_O]

Thanks Mike,

I recall Sony D1 machines cost about $150,000 when they first came out in the mid 80's.

How many computers w/ RR-x could you buy for that? ;-)

 

[jake blackstone]

I don't feel it's my place to quote potential budget figures. But relatively speaking, a telecine room in the "old days" generally cost almost $1,000,000 for the telecine alone. A DaVinci 2K - the most popular color corrector of the HD era - was almost $400K. And a 32 inch Sony CRT monitor - the "Rolls Royce" of that era - was almost $40K new. You also needed clean room environments, an ultrasonic film cleaner with proper ventilation, much more electric power, elaborate signal routing, multiple different videotape formats, and a sizeable vault space - all unnecessary today. Today, the Pablo is among the most expensive - if not the most expensive - of all of the grading systems. So if you substitute the components needed for a DI theater, and compare that to an HD grading suite of the mid to late 90's (and most of them were not using Ranks by then, they were generally using Spirit telecines, but the cost was similar), you're exchanging a 2K for a Pablo, which is something of a wash. You're exchanging a $40K monitor for a projector that's roughly double that. And you're building a bigger room with a large screen. But nowhere are you still incurring the cost of a telecine machine, which was a million dollars all by itself. Nor do you require the costs of all the other items I mentioned. So the answer to your question is that it's far, far less costly today than it was then. But then again, the rates were considerably higher then, the deliverables were more costly (videotape machines were generally around $100K or more each, and tape stock was about $75 per hour), and in general, the business was at least somewhat more profitable, although post production has never been a particularly profitable business. Lower cost of materials and equipment does not always equal healthier or more profitable business models, not when the rates come down severely and not when your customers sometimes become your competitors.

And don't forget the operational costs of the telecine suite. It usually required a full time engineer to be present at all times to just keep the suite operating normally, as for example, when changing from 16mm to 35mm on RANK required a shading operation, unless the company was pretty high end and could afford a separate shading cards for different gages, which cost pretty penny as well. And the cost of RANK CRT tube replacements, which cost a lot and required two engineers to perform. And the cost of the RANK upgrades for something like Metaspeed option or the Digital deflection upgrade. On the other hand, you could keep RANK running for 10-15 years with relatively "inexpensive" upgrades and proper maintenance. And don't forget the huge cost of maintenance contracts, so when your precious DaVinci would go down in the middle of the night, the stuff engineer would call DaVinci and in an hour or so Dwaine or Mark would magically appear and bring your DUI back to life. The list of hidden costs, that doesn't apply any longer with the new technology was simply staggering. Very few companies could afford the costs running proper telecine and had the clientele to support that business plan.

 

[Corey_O]

Don't forget Pin registration gates and way back wet gates. Mag audio machines.

 

[M Most]

And don't forget the operational costs of the telecine suite. It usually required a full time engineer to be present at all times to just keep the suite operating normally, as for example, when changing from 16mm to 35mm on RANK required a shading operation, unless the company was pretty high end and could afford a separate shading cards for different gages, which cost pretty penny as well. And the cost of RANK CRT tube replacements,.....

I didn't want to go back to the standard def era because that would probably be ancient history for many people here. Even I kind of consider it ancient history. So my reply referred primarily to the HD era, which started in about 1998 and in which CCD based Spirit telecines largely replaced tube based Ranks (not completely - there was the C-Reality - but largely), which had the added benefit of minimizing a lot of the maintenance and parts replacement costs long associated with tube based machines. I would also say that Spirits wound up having a similar long shelf life, with far less maintenance costs, but with very similar service contract charges. Of course it should also be said that those support costs were very worthwhile as all three companies (Rank, DaVinci, and Philips/Thomson) did a great job of providing both field service and replacement boards and parts to keep you going.

People who have only been in the business since the mid 2000's have no idea how easy they have it.

 

[Marc Wielage]

A DaVinci 2K - the most popular color corrector of the HD era - was almost $400K. And a 32 inch Sony CRT monitor - the "Rolls Royce" of that era - was almost $40K new.

Doh, it was worse than that if you got all the options, particularly a full-tilt system with two defocus boards and two layer boards, plus options like ColorTrace. I believe we had only one 2K room at Complete with all that, and that DaVinci was well over $500K, including control surface. It's very sobering to consider that the $30,000 Resolve plus a very fast Mac ($10K at most) today would have no problem equalling that system if not doing a lot more.

Today, the Pablo is among the most expensive - if not the most expensive - of all of the grading systems.

I have been told that in the last year, Quantel, Baselight, and Digital Vision have made drastic price cuts in order to stay competitive with Blackmagic/daVinci. I don't even consider Lustre in the game, since Autodesk has kind of shut it down and made it just an add-on to Smoke. It's amazing how much of a reputation Lustre had for a long time; I just had a conversation recently with a talented DP who had the impression that only Lustre had certain features. He was unaware that in 2013, every color-correction system can pretty much do everything, differing more in terms of design than performance.

Lower cost of materials and equipment does not always equal healthier or more profitable business models, not when the rates come down severely and not when your customers sometimes become your competitors. ...People who have only been in the business since the mid 2000's have no idea how easy they have it.

That's the most sobering reality these days. There's a whole generation of post people nowadays who have never worked in film before, which is kind of startling. They take a great deal for granted, and it's sad in some ways, but also enlightening in others.

 

[Frank Cueto]

We managed to keep out Y-Front up and running till 5 years ago, new tube and everything. Indeed those where the days where you needed some form of structure to run the biz. True, people nowadays don't know how ez they have it. BUT, then again, in the midst of running those rooms and being faced with limitations all the time great work came out. People who worked thru that era are a great resource and very knowledgable. Now everything is a website's tutorial away and almost everyone's stuff ends up looking similar.

 

[M Most]

I have been told that in the last year, Quantel, Baselight, and Digital Vision have made drastic price cuts in order to stay competitive with Blackmagic/daVinci. I don't even consider Lustre in the game, since Autodesk has kind of shut it down and made it just an add-on to Smoke. It's amazing how much of a reputation Lustre had for a long time; I just had a conversation recently with a talented DP who had the impression that only Lustre had certain features. He was unaware that in 2013, every color-correction system can pretty much do everything, differing more in terms of design than performance.

What they are all struggling with is finding a good combination of options, alternatives, business models, and price points that can satisfy and hopefully expand their user bases. In the case of Filmlight, they have instituted price cuts, but they've also introduced a less costly but still effective control panel at half the price of the original, as well as software only licensing (Baselight was only sold turnkey for a long time) and various hardware configurations. They have also widened their product line to include on set tools. Digital Vision and Quantel have also adjusted pricing and licensing models. Autodesk has introduced a subscription and term licensing model on Flame Premium that is similar to their licensing options on the 3D products, namely Maya, 3D Studio Max, and Softimage. So yes, they're all looking for what might work. For some market segments, they know that it's pretty difficult to sell against Free. But for others, what they bring to the table is very desirable and the user base is very loyal because of the features and performance. So while Free might be the only choice for the DIY crowd, professionals and facilities who service certain clientele often see it differently. And with regard to feature sets, while all of these products "can" do just about everything, the way they do them is different, both over and under the hood. That means the interfaces and feature sets are different, but it also means that the transforms, math, and other processing that goes on in each of these products can be and is considerably different. If you don't believe me, you might want to have a conversation with Charles Poynton, Josh Pines, or Alex Forsythe about subjects like Resolve vs. Baselight. I can guarantee you it would be surprisingly enlightening....

BTW, just in the name of accuracy, Lustre is currently sold only as part of Flame Premium. Smoke is no longer a part of Flame Premium, although the Smoke timeline, conform toolset, and a number of its features remain as part of the Flame product. Smoke currently exists only as a standalone product on the Mac, and Lustre is not and never has been a part of that product. Also, I believe existing licenses of Lustre on Windows are still supported, but new licenses are not available.

 

[Peter Moretti]

The information in the orginal RAW version is the same, but the interpretation through the gamma curve is different. Since that interpretation is what is presented to you when you start grading, it does have an effect on both how you grade it and how much flexibility you have in doing so. If you use one of the Redgamma curves (any one of them, for the purposes of this discussion) you get a starting image that has been "normalized" into a video like gamma, as opposed to the log coded image that you get with RedlogFilm. Because of this, if your grading involves raising any lower end values, you will likely see more noise working from the Redgamma image than you would from the RedlogFilm approach. That's the nature of how grading works. Color correction is generally a destructive process, and concatenated transforms - particularly those that bring up lower end values - create issues that don't necessarily exist in the original image. Values in the lower end of the scale on a gamma encoded image tend to be noisy because there aren't a lot of levels available for quantization. When those values are transformed into higher ranges, that becomes obvious in the form of random noise. By starting with a log coded image, nearly all of the important values in the image are in a part of the curve where noise isn't a factor. Beyond that, there are the practical considerations of how one goes about grading. I have long advocated using log controls - namely exposure and contrast - as the primary controls for log coded images. In that scenario, you are not separating the image into different brightness ranges for the purposes of balance, you are instead altering the levels of the entire red, green, and blue records all at once. For that to really work as a grading approach, the image really needs to be in log space. So for me, I always prefer interpreting Red files using RedlogFilm because I feel you generally get a better result with log grading with respect to recreating the original photography and the original captured colors with better separation and fidelity. And that, for me, is a much better starting point for any creative interpretation that might or might not be necessary.




Like most professional colorists I know, I rarely use curves, for a number of reasons. Curves can be OK for an overall transform (i.e., converting log images to video or P3 gamma spaces), but unless you really understand what that curve is doing, and what it should be doing, it is far too easy to create artifacts unintentionally by using them. Things such as either "flat" or "too contrasty" faces are common when a curve is not precisely designed or controlled. For this reason, most colorists I know use a predesigned "precise" curve in the form of a LUT for this purpose. On the other hand, my friend Leo Marini (owner of Local Hero Post, and a really talented colorist as well) uses them all the time, primarily to make images "pop" and achieve a brighter, "snappier," and generally "happier" look. You can see this on some of the pictures he's worked on ("Pitch Perfect" is a good example), and for that kind of thing, curves can be useful, and I've sometimes used them for that reason. I will also sometimes use them for opening up the blacks slightly, or increasing midrange contrast. But those are all "touch up" scenarios. As a primary color tool, they are not generally used. For one thing, they cannot easily be controlled by a tactile physical control, like a trackball or a knob. This means that changing their values requires a mouse or pen approach, and for most professional colorists, this largely eliminates the "feel" that is so critical for tactile feedback and brings a lot of the talent of the colorist to the table. By the nature of their design, they also limit you to essentially changing only one thing at a time. For most colorists, grading is always an interactive process that requires the ability to change multiple parameters at the same time, both in terms of the primary color channels and the overall values. With a differential control, such as a trackball, you're never changing only one channel, you're always affecting all three. This, once again, allows you to "feel" your way to the delicate balance that you want. There are times when you might want to change only one channel and in most systems, that ability is also there. But for the most part, you're working in a world of balance, and that is where the "feel" part comes into play. With curves, this ability is lost, and you're constantly moving from one to the other, which is not only inefficient, but leads to inappropriate and misbalanced values. So the short answer to your questions are curves are never my primary control, but they are sometimes useful for specific things.

Mike I want to thank you wholeheartedly for such a complete reply. I very much appreciate the effort you put in to it!

 

[Hrvoje Simic]

Like most professional colorists I know, I rarely use curves, for a number of reasons. Curves can be OK for an overall transform (i.e., converting log images to video or P3 gamma spaces), but unless you really understand what that curve is doing, and what it should be doing, it is far too easy to create artifacts unintentionally by using them. Things such as either "flat" or "too contrasty" faces are common when a curve is not precisely designed or controlled. For this reason, most colorists I know use a predesigned "precise" curve in the form of a LUT for this purpose. On the other hand, my friend Leo Marini (owner of Local Hero Post, and a really talented colorist as well) uses them all the time, primarily to make images "pop" and achieve a brighter, "snappier," and generally "happier" look. You can see this on some of the pictures he's worked on ("Pitch Perfect" is a good example), and for that kind of thing, curves can be useful, and I've sometimes used them for that reason. I will also sometimes use them for opening up the blacks slightly, or increasing midrange contrast. But those are all "touch up" scenarios. As a primary color tool, they are not generally used. For one thing, they cannot easily be controlled by a tactile physical control, like a trackball or a knob. This means that changing their values requires a mouse or pen approach, and for most professional colorists, this largely eliminates the "feel" that is so critical for tactile feedback and brings a lot of the talent of the colorist to the table. By the nature of their design, they also limit you to essentially changing only one thing at a time. For most colorists, grading is always an interactive process that requires the ability to change multiple parameters at the same time, both in terms of the primary color channels and the overall values. With a differential control, such as a trackball, you're never changing only one channel, you're always affecting all three. This, once again, allows you to "feel" your way to the delicate balance that you want. There are times when you might want to change only one channel and in most systems, that ability is also there. But for the most part, you're working in a world of balance, and that is where the "feel" part comes into play. With curves, this ability is lost.

It might be worth adding that potential of curves depends not only on the colorist, but their software implementation as well.

For example, curves in Resolve are poorly implemented, compared to Photoshop and even Color.

They could be controlled by trackballs, with separate trackball allowing control of knee, toe and mid points of the bezier curve, with 3 point curve control. The other option would be having each trackball for each channel (one point curve control). That option would require switching between three points of the curve within each channel. Both options could provide more powerful control compared to mouse or pen control.

There are ways in which they could provide more "feel" but that needs more work done on their implentation and more software developer's interest. I just mentioned one solution, there are others.

 

[M Most]

They could be controlled by trackballs, with separate trackball allowing control of knee, toe and mid points of the bezier curve, with 3 point curve control. The other option would be having each trackball for each channel (one point curve control). That option would require switching between three points of the curve within each channel. Both options could provide more powerful control compared to mouse or pen control.

Well, that kind of implementation is already in Baselight and has been for quite some time. Baselight even has a "magnified" view of the specific area of the curve that surrounds your current control point, and even user selectable spline interpretations. But frankly, I never found myself using curves any more often as a result, nor did I find it any "more powerful" than mouse or pen control. It was a bit more precise (thus making it a more useful "touch up" control), but it didn't change the nature of the curves as single parameter controls, nor did it change the need to look at the curve display as you manipulate them - both of which are efficiency killers for most professional colorists.

 

[Hrvoje Simic]

Well, that kind of implementation is already in Baselight and has been for quite some time. Baselight even has a "magnified" view of the specific area of the curve that surrounds your current control point, and even user selectable spline interpretations. But frankly, I never found myself using curves any more often as a result, nor did I find it any "more powerful" than mouse or pen control. It was a bit more precise (thus making it a more useful "touch up" control), but it didn't change the nature of the curves as single parameter controls, nor did it change the need to look at the curve display as you manipulate them - both of which are efficiency killers for most professional colorists.

Understandable. I agree with the "touch up" as the most common route, if needed.

Regarding Baselight, again...usability of the feature depends on the implementation of the feature, not just the feature itself. Zooming in on the portion of the curve can be counter productive due to: a) additonal redundant clicks; b) limiting perception of the full curve while modifying it; c) misleading into more destructive alterations of the image; d) requiring more focus onto a tool operation and less to its result. All of which can be avoided simply by having a larger curve window. Which requires re-thinking some elements of the UI and some feature implementations. Curves tool is not the only one which would benefit.

Regarding "single parameter", depends on what is the priority. For example, affecting local contrast by shaping a gamma curve can also be more productive or creative in some cases. If used properly and within its limitations. What curves also do is offer intuitive and informative visual feedback about picture alterations, which may help freeing visual memory and mental processes, compared to trackballs, knobs and numerical values. This is generally speaking of course, and excluding individual preferences and developed habits.

Regarding focus on the curve display while manipulating it, the issue is less apparent with larger curve window and its better positioning. Prior factor requires less focus than a tiny window (thread>needle/ball>goal), latter allows usage of peripheral vision. Also, this limitation gets counterbalanced with the previously mentioned benefit of intuitive visual perception of image alterations. Meaning - while one may need to focus a little more to tweak the curve, the mental image of the curve holds many values within itself and less focus may be required later to re-check what was done and establish reference points in some cases. This issue is also a product of UI logic inherited from analog era and hardware. Many things can be improved, while keeping the established and essential requirements. Overlaying curves (and qualifiers for that matter) on top of scopes would be a major step up in many factors.