Steve Yedlin, ASC on Resolution

[Sabovic Adis]

I try to explain,

The projectors have crystal sync just like the cameras. And you can change it. In sweden/stockholm the 35mm Film projectors where set to the standard of 24 fps but a lot of films where actually shot and edited and shown in 25fps in the theaters. And the theaters did not care much and let the setting sit on one of the two no matter frame rate of the film was supposed to be shown in, and you can over / under crank certain projectors quite a bit especially those they use at labs.

Bold#1

Having worked for years as a projectionist, I find that hard to believe. Not once I had to "flip the 24/25 switch" on a projector - never even saw none. But, for example, switching from one projector to another, the so-called changeover, lenses in a 3 lens turret, in turn, I flipped those a lot. :smile5:


Bold#2

That I believe 100%. :rofl:

:thumbsup:

But What Joel says about variable motion blur will be, or is allready possible to certain extent as Graeme explains.
The day we have "infinite frame rates" we can shoot 360 shutter and in post decide what frames out the frame lengths should be mixed into the frame and also with what weight, that way you have the possibility not only to decide if you want a lot or a little motion blur / shutter speed in post. You can aslo decide if you want it sinus wave or sqare etc. And cameras like phantom etc have well enough frame rate to do this allready today, even though it for sure increase the amount of light neeeded.

I don't know...personally, I don't like variables, I prefer constants. Somehow, I find it easier to work when you know upfront what you're dealing with, and I mean not only for "post people" but for DOPs, Directors etc. as well.




E:

@Evin,
Including 16 mm and 70 mm?

 

[jake blackstone]

Some say they prefer to color correct from a DPX file enven if it was generated from RAW file. Do you also prefer to use the DPX?

It's not a matter of preference but a matter of practicality and accepted practices. ALL feature grading that is done at top companies today (and that includes all Technicolor and Delux companies) are done using DPX and EXR. It blows my mind, that after so many years people still cling to the idea, that grading with RAW material is the best way to go.

 

[Björn Benckert]

It's not a matter of preference but a matter of practicality and accepted practices. ALL feature grading that is done at top companies today (and that includes all Technicolor and Delux companies) are done using DPX and EXR. It blows my mind, that after so many years people still cling to the idea, that grading with RAW material is the best way to go.

Not sure how they do it a Technicolor and Delux, but normally the grade is done not as a first step of the post production process but more towards the end. Practice is to set a base light / do small adjustments and then get the material into a container that fits the post / VFX workflow. Then natrually when it´s time for final grade there is usually not that many clips that actually are native camera files any more, retouch, VFX etc simply demands a processing step. So EXR or DPX or what not is used as a Digital intermediate between the raw capture file and what ever the mastring format is in the end.

Then as I understand a lot of colourist tools used do simply not handle all kinds of raw files, so they transcode to something they can work with, even when the above workflow is not the case.

 

[Patrick Tresch]

It blows my mind, that after so many years people still cling to the idea, that grading with RAW material is the best way to go.

RAW R3D 6k or 8k are lighter than any encoded format you mention. So when the pipeline is calibrated and there are no FX (perhaps not the films you do) there is no advantage to grade through EXR or DPX.

 

[Marcus Friedlander]

RAW R3D 6k or 8k are lighter than any encoded format you mention. So when the pipeline is calibrated and there are no FX (perhaps not the films you do) there is no advantage to grade through EXR or DPX.

Even for films that have some VFX, but not a ton, it doesn't make sense to transcode the entire project to EXR or DPX.

The movies that I've shot and graded only have a couple dozen VFX shots top. So it makes more sense to just transcode those shots, but keep the rest in R3D.

 

[Marcus Friedlander]

Some add-on fuel to the discussion :tongue_smilie:

https://ascmag.com/blog/presidents-desk/president-039-s-desk-resolution-resolved

That was a great read! Thanks for sharing!

 

[Christoffer Glans]

With IPP2 there should be no problem working with R3Ds when grading and pull out DPXs in LOG3G10 or LOG3G12 with the same output LUT to check images while doing VFX. Not using RAW is missing out on things like HDRx and I've had no problems getting past the motion artifacts when using that feature. I know a lot of colorists say that they can pull off higher looking DR within the normal range of an image, or use ISO splits, but for some shots it's not even close comparing something shot in 19-22 stops with normal shots. It's the most underused feature with Red cameras but getting accustomed to it really opens up a lot, especially for uncontrolled lighting situations. Just keep that shutter fast and it works great after post.

 

[Marc Wielage]

It's not a matter of preference but a matter of practicality and accepted practices. ALL feature grading that is done at top companies today (and that includes all Technicolor and Delux companies) are done using DPX and EXR. It blows my mind, that after so many years people still cling to the idea, that grading with RAW material is the best way to go.

Jake is 100% right. What is critical is that when they create the DPX or EXR transcoded files, they have to use optimized settings from the Raw files to make sure it's done correctly. But the assistants and conform artists at Tech and Deluxe are first-class and understand what to do.

I have warned filmmakers before even when we do use the Raw files, it's much, much too time-consuming to do a lot of tweaking within the Raw panel. As long as it's set to some kind of reasonable base grade, we can do the rest easily from RedLogFilm and get what they need.

Even for films that have some VFX, but not a ton, it doesn't make sense to transcode the entire project to EXR or DPX. The movies that I've shot and graded only have a couple dozen VFX shots top. So it makes more sense to just transcode those shots, but keep the rest in R3D.

It's generally not the way it's done for A-list features and network TV series in NY and LA. I have done indie films where it was a combination of R3D files and DPX files (or R3D and ProRes 444 4K files), and I defy anybody to tell the difference. In my case, as long as the system is up to snuff, I don't have a problem using the Raw files provided there's time. If we're in a huge hurry, then I will convert to ProRes444 just to move it along faster and provide less of a load on the computer. It's all still at 4K, and in some higher-res cases, I have converted to the original 5K or 6K aspect ratio just to get that extra bit of resolution in the event of blow-ups.

 

[Phil Holland]

This is 100% true. What is critical is that when they create the DPX or EXR files, they have to use optimized settings from the Raw files to make sure it's done correctly. But the guys at Tech and Deluxe are first-class and understand what to do.

I have warned filmmakers before even when we do use the Raw files, it's much, much too time-consuming to do a lot of tweaking within the Raw panel. As long as it's set to some kind of reasonable base grade, we can do the rest easily from RedLogFilm and get what they need.

Lot's of workflows out there these days and of course depending on the pipe, but....

My general practice for years on features delivering plates or shots has been:
- in REDCINE-X Pro dial in the desired Color Temperature and Tint as well as ISO Rating
- Export desired Color Space and Gamma Curve (these days and pretty much for the last year+ REDWideGamutRGB and log3G10)
- Export DPX to desired output resolution defined by the studio or production.
- Include Quicktimes w/ generally a REC.709 output as well.
- SHIP DRIVE AS FAST AS HUMANLY POSSIBLE

I also include the .R3Ds if they want them and also include a nice zip file of the newer Output Transforms as some houses have yet to tackle this specific workflow.

 

[jake blackstone]

Even though I specifically stated, that "ALL feature grading that is done at top companies today (and that includes all Technicolor and Delux companies) are done using DPX and EXR", many posters immediately started to talk about how THEY prefer to grade with RAW. It really doesn't matter how you THINK it should be done. It is done with transcoded material because this workflow had been thoroughly tested, perfected, optimized and after many years this workflow is a generally accepted by virtually all top feature companies. And, yes, virtually ALL shots, that are supplied for VFX work are PRE-GRADED. The last three films, that are the winners for Best picture for cinematography, were graded by Steve Scott at Technicolor on Lustre, which is perfectly optimized for working with DPX, but is downright dreadful when working with any other types of compressed images like R3D.

 

[Björn Benckert]

Jake is 100% right. What is critical is that when they create the DPX or EXR transcoded files, they have to use optimized settings from the Raw files to make sure it's done correctly. But the assistants and conform artists at Tech and Deluxe are first-class and understand what to do.

I have warned filmmakers before even when we do use the Raw files, it's much, much too time-consuming to do a lot of tweaking within the Raw panel. As long as it's set to some kind of reasonable base grade, we can do the rest easily from RedLogFilm and get what they need.


It's generally not the way it's done for A-list features and network TV series in NY and LA. I have done indie films where it was a combination of R3D files and DPX files (or R3D and ProRes 444 4K files), and I defy anybody to tell the difference. In my case, as long as the system is up to snuff, I don't have a problem using the Raw files provided there's time. If we're in a huge hurry, then I will convert to ProRes444 just to move it along faster and provide less of a load on the computer. It's all still at 4K, and in some higher-res cases, I have converted to the original 5K or 6K aspect ratio just to get that extra bit of resolution in the event of blow-ups.

If you master at 4k then I think most computers rather chew on the raw files than 16Bit DPX. We now did the finnishing on 10x 45 minute 7k raw mastered in 4k. All done on a couple of trashcans. No problems what so ever. debayer is set to a quarter while working and then we rendered at full res. All from media managed / trimmed r3d's. VFX clips we did from 10bit 4k DPX files and the we keep those on a top layer in the resolve time line and as its not so many VFX shots we could simply render everything out again as 7k or even up scaled to 8k to get such master. Doing this whit DPBX file s would not be possible for us the machines would not like that amount of DPX files and especially not if they where 7k.

And most of all to do a pregrade transcode to r3d trim is way faster, undistructve and takes less space than doing it to DPX. So as I see it if your grading tool can work with the r3d's as resolve can. Just work with your image scaling to make it run smooth while grading then that approach is what I recommend.

 

[Hrvoje Simic]

And most of all to do a pregrade transcode to r3d trim is way faster, undistructve and takes less space than doing it to DPX. So as I see it if your grading tool can work with the r3d's as resolve can. Just work with your image scaling to make it run smooth while grading then that approach is what I recommend.

Every route has advantages and drawbacks.

DPX/EXR route bombs the disk space and data management but helps to assure proper RAW development and consistency.
RAW route kept reduces disk space and leaves more room for changes but more room for mistakes as well.
And more complex productions tend to avoid those.

 

[Christoffer Glans]

RAW route kept reduces disk space and leaves more room for changes but more room for mistakes as well.
And more complex productions tend to avoid those.

That's just a matter of routine. If you got a good IPP2 routine with one-lights and proper output-LUTs you won't get into trouble. The only thing that change between DPX and RAW is that you can go back and change the RAW settings. Worst thing if you go DPX is that all the one-light mistakes are fixed in place and that's a far worse problem then having RAW controls. I don't really see what the problem is working with RAW. Everyone talk about how much more complex it is and inviting mistakes, but there's no difference really. Whenever you do VFX, you just lock the one-light and do DPX for those shots. It saves disk space, is faster and let's you change something if any problems occur.

I've never had any problems with RAW workflow and it's never been really that different from times I grade DPX or ProRes files. Only the GPU/CPU heavy debayer process is the difference for me.

 

[Josh Wilkinson]

Just really watched this for the first time after hearing about it.

I'm seeing pretty obvious clarity differences between different images at 4k. Maybe my eyes are tricking me. It was especially obvious when we was switching between the 6k and the 2k images in part 2.

 

[Marc Wielage]

Just really watched this for the first time after hearing about it. I'm seeing pretty obvious clarity differences between different images at 4k. Maybe my eyes are tricking me. It was especially obvious when we was switching between the 6k and the 2k images in part 2.

I just spoke to somebody who was at the LA SMPTE demo yesterday, and he told me there was no difference. Yedlin made a very strong point at the demo.

Here's a link to the two demos in which Yedlin presents his case:

http://yedlin.net/ResDemo/

I think Mr. Yedlin is a very bright guy and I wouldn't be too quick to call him wrong. He clearly has thought out his argument very well, and he's been around for decades, shooting both film and digital.

 

[Josh Wilkinson]

I just spoke to somebody who was at the LA SMPTE demo yesterday, and he told me there was no difference. Yedlin made a very strong point at the demo.

Here's a link to the two demos in which Yedlin presents his case:

http://yedlin.net/ResDemo/

I think Mr. Yedlin is a very bright guy and I wouldn't be too quick to call him wrong. He clearly has thought out his argument very well, and he's been around for decades, shooting both film and digital.

I wasn't necessarily saying he was wrong. Resolution for resolution sake is pretty silly for the most part and I definitely agree the pictures out of anything above $15k are just phenomenal.

I think his argument is a little shortsighted/simple (maybe intentionally so). I think a lot of his argument is based around people viewing in a theater or small relatively poor quality 4k displays, which has and is changing every year. (affordable 4k short throw laser projectors come to mind as being maybe the most optimal way of viewing upcoming content.)
Also doesn't evaluate the package it comes in or the tech behind each.
Very much reminds me of the "gear doesn't matter" saying.

I have a 43in 4k calibrated monitor setup at optimal viewing distance (about 36 inches) and I can see distinct minor differences between each image. While I don't think the average viewer would be able to point them out, I think sub consciously they are noticeable.

I definitely enjoy Yedlins work in general. And its not like I just shot of most successful movie of all time either. Just my 2 cents.

 

[Marcus Friedlander]

...It was especially obvious when we was switching between the 6k and the 2k images in part 2.

I was at that LA SMPTE demo and the difference between the two images was so microscopic that it literally caused the audience to break out in laughter.

But I believe that came down more to his control of algorithms, rather than an inherint lack of value in higher resolution sampling. As it's important to note that it's still an A65 source that he's starting from. I'm sure the test would be not quite as favorable if he used the 6k film scan, that's not actually resolving 6k, as the source.

All of this showed to me that we should be talking less about resolution, and more about alograthims. Because what his demonstration clearly showed is that if you want authorship of your image, you need to be able to control the various algorithms used to interpret your footage (from his discussion I noted: grain, halation, sharpening, downscaling, upscaling, and noise reduction)

I think it's also important to remember that all of his arguments are contingent on viewing angles relative to screen size and viewing distance. From the research he showed, your 3 foot viewing angle for your 4k monitor is far outside the norm of where the average consumer sits relative to their screen.

Also, his focus is specifically on cinema presentation, and the viewing angles that relate to it. So, after seeing his presentation in a cinema, I have to say that I'm a believer. Algorathims matter more than resolution in determining final spacial fidelity (provided you've started with a good enough source. Which means pretty much any camera made in the last 5 or so years.)

 

[James Lee Marsh]

Just checked this out and its great. Great work. I wish there was a easy and fast way to educate clients and producers about this. Im still holding onto my old and trustworthy epic-x because of the reasoning talked about in this presentation.

What is the node program you are using in part 2?

 

[Kenneth C Merrill]

I finally got around to watching all of this. Really incredible, well-articulated work by Yedlin. Someone should buy that guy a drink!

My thoughts:

1. As with everything image-related, some of the conclusions that Yedlin comes to are subjective. One of his key premises is that at 3K we already surpassed the image resolution needed for true "spacial fidelity" in most applications--and so making pixels smaller and adding more pixel resolution does not enhance the overall experience of the film. At least not as much as other factors such as halation, scaling, etc.

Ironically his images drew me to completely the opposite conclusion. While the spacial fidelity of the cameras does not correlate directly with their pixel resolution, with the exception of the film scans, the higher resolution images tended to have more visible spatial fidelity, and the difference between the 3K Alexa XT and the 6K Alexa 65 was just incredible. I honestly don't know how you can look at those two images side by side and then say pixel resolution doesn't make a difference.

Just speaking from my own experience, contrary to what some people can say, I CAN tell the difference between 4K and 2K imagery. Heck, I even edited some Alexa 2K files recently and was surprised at how irresolute they look--my eyes have slowly become accustomed to 3K, 4K+ footage.

2. I disagree with his methodology at the beginning of Part 2. I understand that he is trying to show simply the technical difference between a 2K->4K vs 6K->4K scaled image, but the problem is the 2K image he is using is originally scaled from 6K! While this works fine as a sort of mathematical demonstration, it doesn't reflect real-world results when you compare a 2K capture camera vs a 6K capture camera. You can see this later when he compares the XT with the 65, the differences between the 3K and 6K images are much more pronounced than suggested here at the beginning.

3. He used many different lenses for the test, to accommodate different sensor sizes and such. This is yet another factor that can affect spacial fidelity and may account for some of the more irresolute images in the higher K cameras, particularly the IMAX 11K scans.

4. I used to believe that 35mm film had 4K+ worth of resolution, but I think Yedlin has shown it tops out at 3K max, though the difficulty here is not necessarily capturing the spacial fidelity but the color fidelity of the film.

5. I cannot agree more with his concerns about compression. Especially as more of our viewing is content that is streamed from an online source, it will be challenging to view 4K+ content in a way that isn't destroyed by compression.

6. I think Yedlin's overall premise is correct. It's not JUST about how many K's your camera has, and I do agree that there is a limit to the amount of resolution that our eyes can see. I disagree that we passed that limit at 3K. Based on what I've seen from current imaging technology, up to 8K (Red and Canon), I think the visual limit of perception probably lies somewhere around 5 or 6K. I also think Yedlin is right that other considerations are often more impactful than pixel count when it comes to the audience's perception of an image--but I also don't think it's a zero sum game. As technology marches forward, we will have high resolution images with good color fidelity and broad dynamic range, and discussions like this will probably become antiquated to some extent.

 

[Matt Bai]

I finally got around to watching all of this. Really incredible, well-articulated work by Yedlin. Someone should buy that guy a drink!

My thoughts:

1. As with everything image-related, some of the conclusions that Yedlin comes to are subjective. One of his key premises is that at 3K we already surpassed the image resolution needed for true "spacial fidelity" in most applications--and so making pixels smaller and adding more pixel resolution does not enhance the overall experience of the film. At least not as much as other factors such as halation, scaling, etc.

Ironically his images drew me to completely the opposite conclusion. While the spacial fidelity of the cameras does not correlate directly with their pixel resolution, with the exception of the film scans, the higher resolution images tended to have more visible spatial fidelity, and the difference between the 3K Alexa XT and the 6K Alexa 65 was just incredible. I honestly don't know how you can look at those two images side by side and then say pixel resolution doesn't make a difference.

Just speaking from my own experience, contrary to what some people can say, I CAN tell the difference between 4K and 2K imagery. Heck, I even edited some Alexa 2K files recently and was surprised at how irresolute they look--my eyes have slowly become accustomed to 3K, 4K+ footage.

2. I disagree with his methodology at the beginning of Part 2. I understand that he is trying to show simply the technical difference between a 2K->4K vs 6K->4K scaled image, but the problem is the 2K image he is using is originally scaled from 6K! While this works fine as a sort of mathematical demonstration, it doesn't reflect real-world results when you compare a 2K capture camera vs a 6K capture camera. You can see this later when he compares the XT with the 65, the differences between the 3K and 6K images are much more pronounced than suggested here at the beginning.

3. He used many different lenses for the test, to accommodate different sensor sizes and such. This is yet another factor that can affect spacial fidelity and may account for some of the more irresolute images in the higher K cameras, particularly the IMAX 11K scans.

4. I used to believe that 35mm film had 4K+ worth of resolution, but I think Yedlin has shown it tops out at 3K max, though the difficulty here is not necessarily capturing the spacial fidelity but the color fidelity of the film.

5. I cannot agree more with his concerns about compression. Especially as more of our viewing is content that is streamed from an online source, it will be challenging to view 4K+ content in a way that isn't destroyed by compression.

6. I think Yedlin's overall premise is correct. It's not JUST about how many K's your camera has, and I do agree that there is a limit to the amount of resolution that our eyes can see. I disagree that we passed that limit at 3K. Based on what I've seen from current imaging technology, up to 8K (Red and Canon), I think the visual limit of perception probably lies somewhere around 5 or 6K. I also think Yedlin is right that other considerations are often more impactful than pixel count when it comes to the audience's perception of an image--but I also don't think it's a zero sum game. As technology marches forward, we will have high resolution images with good color fidelity and broad dynamic range, and discussions like this will probably become antiquated to some extent.

:cheers2: