Roger Deakins, ASC vs. 4K

[Scott Crawley]

Any section or subsection on a Red operated forum, whether public or private is going to have a Red bias by design. They like it that way, and keep it that way. If you want frank and open, a neutral venue is the only place to go.

David Mullen contributes to a number of sites like you describe. Google him and you will find them.

 

[Blair S. Paulsen]

Feud for thought...

Feud for thought...

On the simplest level you have two companies making products for the same market who employ multiple strategies to spur sales. While the battle may start with actual empirical product differentiation, the descent into tactical marketing-speak tends to start about one second later. Hardly news there. My point is that while such a "Holy War" may lead to frustrating obfuscation, its endemic.

So, how do we punch through the "noise" to have a productive discussion about actual image metrics and real world functionality?

James feels like RedUser has critical limitations as a forum for exploring the evolving universe of digital cinematography. No offense to our community here, but to paraphrase the philosopher Popeye, we are what we are - which is a group of people pre-disposed to appreciate particular aspects of the RED cams. Like politics, true neutrality is so rare that its not really a factor.

FWIW I know James and I believe he is honestly trying to push digital cinematography forward. He wants to deepen his understanding of the science and the art. Someone with those goals will always find an ally in me. I'd like to believe that even on "RedUser" we can delve into the process of getting intended results from digital devices by drilling into the messy details. IMO it starts with acknowledging that different design choices, different components, etc yield different results and as cinematographers we have a lot to gain by figuring out ways to take advantage of the strengths and mitigate the weaknesses of any digital camera.

Cheers - #19

 

[Jon MIchael Puntervold]

On a closing note: Competition is good!

Right no, neither the Red Epic, nor the Alexa is a perfect camera in my opinion.

Epic MX still has some issues on the highlight roll-off, and Alexa is not shooting raw without a cumbersome external system.

My guess is that Red Epic Dragon will feature a real evolution in Red image quality and organic/natural "feel" (in a lack of a better word), and my guess is that the next generation Arri cameras, succeeding the Alexa will feature on-board Raw encoding.

Red and Arri are pushing the envelope in digital cinematography. I credit mr. Jarred and his team for the revolution, and I will probably always remain loyal to Red because of it. But Arri offers good and valuable competition, maybe due to their heritage aswell?

Neither the less. Fanboys are annoying on both sides of the fences. I get _really_ angry when DoPs right from filmschool claims that Red is just shit. Who the hell are they to judge the work of our greatest filmmakers on the international scene - and claim they shoot on shit - just because one of their institutionalised mentors told them so - based on nothing but superstition of the new and unfamiliar? Likewise, people at the Red camp claiming that i.e. Deakins is delusional because of his love of the Alexa are just as unserious and annoying...

 

[Vadim Bobkovsky]

Love the quote James, and yes, I think the Arri Alexa is a fine camera, it's just not for me. And I think that is what separates a lot of Alexa owners I talk to and RED owners. I see the benefit of the Alexa and comfort to those who like the company and camera. Arri/Alexa folks generally are divisive and can't see a place for RED anywhere. It's silly. And Vadim... I respect Mr. Deakins about as much as anyone. He has created outstanding work my point is... So has RED.

I have some private thoughts about the idea of 2K being "good enough" for major features (especially in IMAX). I'm not posting it here on Reduser, because I respect guys from Red. I admire the work of master cinematographers, shooting movies on whatever medium they want. I'm not booing Deakins or anyone else, I love filmic quality and I understand where he comes from in terms of colors and pleasing "organic" picture characteristics. One of my friends generously let me to mess with a Sony 4K camera and recorder for a weekend outside the city, later while I've been playing with the raw data on my PC, I was amazed how much the picture being pushed beyond reason and common sense still had that pleasing, dare i say filmic grain, instead of typical harsh noise in underexposed areas and in higher ISO modes. As I compared those shots to 4K film scans I though to myself - wow, I can't even tell what is what by how the grain looks anymore. I'm sure professionals still see the differences right away, but that moment was some kind of revelation to me personally. That being said, while 2K on a small screen is OK for me, I can't help myself when I sit in front of a giant-ass IMAX movie screen and the picture just isn't sharp. It's still Bond, James Bond, but my God, I'd prefer to watch way sharper and brighter pictures on those type of screens, to the point of buying ticket to the smaller screen next time. Taking Chris Bale's words - "It's f...ing distracting!".

It's still September, but I can't wait to see the first frames from a mighty Dragon. I guess one of them already named "Smaug". Obviously .

 

[James T Mather]

Thanks for the positive vote of confidence Blair. Nice post outside of that too. We're all trying to get a deeper understanding of this ever-evolving technology we find ourselves immersed in. I look forward to the Dragon Sensor and what opportunities it brings. I hope that aspects like "organic image feel", noise characteristics and roll-off at the top-end are considered - if so, then posting awkward questions is worth it (not that I enjoy making anyone's day more uncomfortable). I feel sure that Mr. Nattress with his alchemy will bring new exciting developments in colour science too. There is no doubt that on paper the Red Epic kicks the competition squarely in the imponderables - however, the battle lines are shifting and it may be time to explore other aspects of the capture process.

Meanwhile, I'm settling down to watching Mr. Pfister's very fine work on The Prestige. Great script, score, editing, art department, music and directing too. The talented bastards

 

[Blair S. Paulsen]

In the interest of trying to separate the facile descriptors from the hard evidence lets talk about the perception that the Alexa images look more filmic.

Film, in all its well earned glory, is still just a reproduction of reality that cannot escape the parameters of its own physical manifestation.

Over time, due to advancements in chemistry and manufacturing processes (not to mention how the makers of film stocks responded to a marketplace whose predilections morphed with the art form) the look of film itself is hardly static. Add in the relationship between various production techniques and the medium, post processes, etc and you have quite a "brew" on your way to the image on the screen.

My point is that for the viewer, especially the sophisticated viewer ( a catagory that I daresay Mr. Deakins and most of us fall into), we bring a reference library of built expectations with us when we evaluate moving images. This is a roundabout way to make the point that IMO Arri chose to imbed what I would consider a film LUT into the Alexa's internal processing pipeline. If, in fact, I am correct about this, it points to a basic philosophical divergence between RED and Arri.

At the risk of over simplifying (stop laughing) Arri's goal was a digital camera that made images that looked like Kodak 5219 stock. Whatever non-linear hue characteristics, density distribution, etc that Arri could mimic they did. Then they did something really smart, instead of giving the user an offline proxy for editorial and a heavyweight version for conform/finish they took a reasonably modern codec (ProRes) and used it as a container for Log-C which is essentially the same log curve used for DPX film scans. Now they had a robust file in a tidy package that could serve both creative editorial and finishing operations with acceptable levels of compromise for most projects.

RED, OTOH, wanted to give the user the power to decide whether they wanted their imagery to look like Kodak 5219 - or something completely different. Depending on the project, your working style, production eco-system, etc this can be either liberating or fraught with peril. For me, and many other RedUsers, the processing of the RAW data set is simply the next step in crafting the image sought. If so desired I can always drop in a film LUT if that suits me - but it hasn't been done for me. Yes, there is an ArriRAW option for the Alexa that offers more control but at a significant cost penalty. RED's compact SSD mags, clever RedCode compression and Red Station readers are significantly more user friendly than the Codex piggybacks, transfer stations and mini-SAS I/O that has made ArriRAW so rarely utilized.

IAC, I prefer the data load to image depth equation of the RedCode RAW workflow vs either the limitations of the ProRes route or the overhead of the ArriRAW path - but that's just one person's preference.

Based on discussions with elite colorists, the other big piece of this "filmic" perception issue is the color matrix. The 10 bit Log-C files in the ProRes container the Alexa creates share a number of attributes with film scans. I contend that for many years the fundamental design philosophies of the color grading tools in our industry (like the DaVinci, Scratch, Baselight, Nucoda, etc) have been optimized for 10 bit log DPX film scans. The result is that colorists can achieve the looks they are being asked for in familiar and efficient ways - especially if the decision maker in the room is seeking the familiar look and feel of film. Moreover, since actual delivered resolution in a traditional film pipeline is rather low (see other threads for details) the lower resolution of the Alexa imagery likely matches the preconceived target in the minds eye. So in addition to a familiar color palette, perhaps "too much" resolution is simply not congruent with many DP's sense of the intended or optimum image...

Cheers - #19

 

[Marc Wielage]

I think the only issue I have with this forum is that it's public and forces people to come out strongly in favour of one side or the other (usually one side), like myopically supporting a football team Alexa/Red, Liverpool/Man United, Blur/Oasis, Mac/PC, My religion/Your religion. This drowns out relevant discussion.

I agree completely, and that was part of my reason for the discussion.

Doesn't everyone want to have a serious open discussion as to why the Alexa is making such inroads into feature film and Television production - or are all those Alexa people just "uninformed simpletons"?

Yes -- that to me was the most intriguing comment in the article. The danger is that the American TV industry sometimes goes along a "flavor of the week" mentality, where X format is king this year, and next year it's Y. I can remember a time in the 1980s when people said "film sitcoms are dead," because of the success of Cosby, Family Ties, and similar shows. Suddenly, Seinfeld and Friends brought film back, and it was huge for almost 10 years. Then HD came in and blew film away. Then sitcoms went away, and hour-long dramas came in. And then, reality shows slammed everything.

So things do come and go in cycles. What's important to remember is, whether it's 4K or not doesn't matter -- the real issues are how reliable the camera is, how well it works for production, how it fits with workflow, and so on. I think we're about to be deluged with a glut of crappy 4K cameras, to the point where the K really won't matter. If you have a college kid with a $5000 4K camera, is he going to be able to compete with an experienced DP with an Epic? It'd be sad to make that decision solely on the basis of price and resolution. It's all the other stuff that's really important.

 

[Richard Goodwin]

In the third part of the interview Roger Deakins says specifically he did not know there was going to be an IMAX release until after he had chosen the Alexa (http://www.youtube.com/watch?v=00OGyWADHyM&feature=relmfu). So, its not accurate to critique the IMAX uprez as if that was a factor in his choosing Alexa; or that the artifacts of an uprez was something he had planned on.

 

[Marc Wielage]

The 10 bit Log-C files in the ProRes container the Alexa creates share a number of attributes with film scans. I contend that for many years the fundamental design philosophies of the color grading tools in our industry (like the DaVinci, Scratch, Baselight, Nucoda, etc) have been optimized for 10 bit log DPX film scans. The result is that colorists can achieve the looks they are being asked for in familiar and efficient ways - especially if the decision maker in the room is seeking the familiar look and feel of film.

I would agree to a point. The only exception is that not everybody does film log scans the same way. I worked on a project some years ago that had scans from eFilm, ILM, Stan Winston, and Technicolor, and all four scans were very different. I could still push the color correction where I needed it to go, but it took some effort, simply because they were so far apart.

I have also had projects with Alexa and film in the same scene (typically with film for slo-mo), and generally had no problem getting them to match. Haven't had film and Red in the same scene. We did have a show (the WB's Supernatural) where for part of a season, I believe they had Red and Arri D21 in the same scene, and that was a struggle. I'm a big believer in not mixing cameras when possible, but I accept that sometimes, it's not avoidable.

RedLogFilm looks fine to me, and I think Red took a lot of time and trouble to arrive at those settings.

 

[Joe G.]

"d) human texture and color (Arri modeled their sensor off Vision 3, Red repurposed an industrial sensor)"

The look is more pleasing, and perhaps this explains it. I happen to think the Blackmagic sensor is also more pleasing.

Waiting to see what the Dragon sensor offers. This may change Deakins, and a lot of other people's, mind on the matter.

 

[Blair S. Paulsen]

Whatever the provenance of the MX sensor, or the Arri sensor for that matter, I am far from convinced that is the key differentiator (though it could be a bigger factor than I suspect and due to obvious competitive issues I wouldn't expect either manufacturer to break it all down for us ;-) FWIW, if I took a poll of the most sophisticated people in this industry who regularly work with both Alexa and Epic/Scarlet footage I think the top ranked response would be that its the color matrix that most defines the default look of each camera.

IAC, for obvious reasons, experienced talent (like Roger Deakins) has incentive to utilize a process they know well and that gives them a high degree of confidence that the final image will look the way they intended. That said, I think its illuminating that top camera teams like the ones on Spiderman, Prometheus, etc who had the time and budget to do extensive testing came to appreciate the Epic platform more and more as they gained deeper understanding of its capabilities.

Cheers - #19

 

[Clint Lealos]

I'd like to have coffee with James and Blair sometime...

Actually, reading all these level headed and thoughtful comments is refreshing

 

[Karim D. Ghantous]

I have opinions on almost everything so I may as well chime in: if the Alexa were smaller and cheaper than the Scarlet, which it should be considering its technology, then I'd love to have one.

It's pretty fair to say that nobody dislikes the Alexa's image. Let's dispense with that strawman. The problem is that thanks to ignorance and urban myths (ZOMG it overheats!!111!11one11eleven) productions are not always making the best choices. IMHO.

But hey, this is a forum. Roger Deakins owes me nothing and likewise the other way. But the worse thing is when people feel that opinions shouldn't be expressed. Let's be glad that they can be.

 

[Ian Vertovec]

Epic MX still has some issues on the highlight roll-off, ...

It should be noted that you cannot discuss highlight roll-off separate from dynamic range. Since all digital sensors are linear, any perception of roll-off is a function of the curves being used to encode/decode the data. If you want to improve Epics highlight roll-off, use different curves.

 

[Ian Vertovec]

the top ranked response would be that its the color matrix that most defines the default look of each camera.

That's also the easiest thing to change about a camera's look. Resolution, dynamic range, and bit depth cannot ever be added if it wasn't there to begin with.

 

[James T Mather]

In the interest of trying to separate the facile descriptors from the hard evidence lets talk about the perception that the Alexa images look more filmic.

Film, in all its well earned glory, is still just a reproduction of reality that cannot escape the parameters of its own physical manifestation.

Over time, due to advancements in chemistry and manufacturing processes (not to mention how the makers of film stocks responded to a marketplace whose predilections morphed with the art form) the look of film itself is hardly static. Add in the relationship between various production techniques and the medium, post processes, etc and you have quite a "brew" on your way to the image on the screen.

My point is that for the viewer, especially the sophisticated viewer ( a catagory that I daresay Mr. Deakins and most of us fall into), we bring a reference library of built expectations with us when we evaluate moving images. This is a roundabout way to make the point that IMO Arri chose to imbed what I would consider a film LUT into the Alexa's internal processing pipeline. If, in fact, I am correct about this, it points to a basic philosophical divergence between RED and Arri.

At the risk of over simplifying (stop laughing) Arri's goal was a digital camera that made images that looked like Kodak 5219 stock. Whatever non-linear hue characteristics, density distribution, etc that Arri could mimic they did. Then they did something really smart, instead of giving the user an offline proxy for editorial and a heavyweight version for conform/finish they took a reasonably modern codec (ProRes) and used it as a container for Log-C which is essentially the same log curve used for DPX film scans. Now they had a robust file in a tidy package that could serve both creative editorial and finishing operations with acceptable levels of compromise for most projects.

RED, OTOH, wanted to give the user the power to decide whether they wanted their imagery to look like Kodak 5219 - or something completely different. Depending on the project, your working style, production eco-system, etc this can be either liberating or fraught with peril. For me, and many other RedUsers, the processing of the RAW data set is simply the next step in crafting the image sought. If so desired I can always drop in a film LUT if that suits me - but it hasn't been done for me. Yes, there is an ArriRAW option for the Alexa that offers more control but at a significant cost penalty. RED's compact SSD mags, clever RedCode compression and Red Station readers are significantly more user friendly than the Codex piggybacks, transfer stations and mini-SAS I/O that has made ArriRAW so rarely utilized.

IAC, I prefer the data load to image depth equation of the RedCode RAW workflow vs either the limitations of the ProRes route or the overhead of the ArriRAW path - but that's just one person's preference.

Based on discussions with elite colorists, the other big piece of this "filmic" perception issue is the color matrix. The 10 bit Log-C files in the ProRes container the Alexa creates share a number of attributes with film scans. I contend that for many years the fundamental design philosophies of the color grading tools in our industry (like the DaVinci, Scratch, Baselight, Nucoda, etc) have been optimized for 10 bit log DPX film scans. The result is that colorists can achieve the looks they are being asked for in familiar and efficient ways - especially if the decision maker in the room is seeking the familiar look and feel of film. Moreover, since actual delivered resolution in a traditional film pipeline is rather low (see other threads for details) the lower resolution of the Alexa imagery likely matches the preconceived target in the minds eye. So in addition to a familiar color palette, perhaps "too much" resolution is simply not congruent with many DP's sense of the intended or optimum image...

Cheers - #19

All very true plus one of two things I did not know in there - great post. Nice one.

 

[James T Mather]

Great new info and an illuminating discussion to boot. Thanks all. Very interesting. I hope the boys at Red are considering film emulation on the Dragon - at the very least to generally match it's characteristics.

 

[Blair S. Paulsen]

(Note: this was in response to my note about the color matrix) That's also the easiest thing to change about a camera's look. Resolution, dynamic range, and bit depth cannot ever be added if it wasn't there to begin with.

I defer to your far greater experience as a colorist and agree that the color values can be easily adjusted. That said, I still contend that where the color starts influences perception (what was seen on the monitor), and in a collaborative art like ours establishes a common point of reference for color that its safer to stick with. I realize that a director, producer or DP may have the juice to push the look anyway but it does create a bit of a speed bump. Just sayin'

Cheers - #19

 

[Ian Vertovec]

I defer to your far greater experience as a colorist and agree that the color values can be easily adjusted. That said, I still contend that where the color starts influences perception (what was seen on the monitor), and in a collaborative art like ours establishes a common point of reference for color that its safer to stick with. I realize that a director, producer or DP may have the juice to push the look anyway but it does create a bit of a speed bump. Just sayin'

Cheers - #19

I totally agree!

 

[Blair S. Paulsen]

T. It should be noted that you cannot discuss highlight roll-off separate from dynamic range. Since all digital sensors are linear, any perception of roll-off is a function of the curves being used to encode/decode the data. If you want to improve Epics highlight roll-off, use different curves.

Thank you Ian. This is a key point that I believe many experienced but less technically oriented folks don't completely wrap their heads around. A natively log medium like film has some very different math embedded in the relationship between densities vs a linear device like a digital cinema camera.

I would even go so far as to suggest that grading toolsets honed over the years for best results with log sources often fail to do justice to the linear scale. If I have that wrong I would love to be taken to school by someone who can explain why. Seriously, I really want to know. I have had some smart cookies tell me that the reason that some of the best looking RED shot material went to log DPX files via RedLogFilm was not just because there were VFX elements involved or because it simplified DI, but because it made it easier to achieve the look being asked for in the bay. They assert log grading techniques get better results faster.

BTW, I carefully manage the shape of the shoulder in the curves to manage roll off rendition on many of my images and consider it one of the most meaningful steps in RAW development. YMMV.

Cheers - #19