Post 4k screening and Red picture issues

[Iron Possum]

After seeing the short film "Crossing the Line" on a 4k projector in Sydney last night I have a little feedback, which are my first impressions of seeing the Red One footage up close and personal. Unfortunately I didn't get to ask the following questions of Ted whilst at the screening.

The pictures are indeed amazing, especially for this price point, no arguments about that 4k resolution from me at all. However I have to say that I was somewhat disappointed by the apparent colour fringing around some of the highlights and the way they clipped, especially in some of the cloud areas. It struck me as quite a harsh look as opposed to the way film deals with these areas of the shoulder, perhaps better than the average video camera but still video like.

I also felt that skin tones in some situations seemed to take on that all to familiar plastic video like texture that you just dont see in film. I'm wondering if these skin tones are a symptom of a knee point set to low and an aggressive knee slope that has reduced the tonality and gamma of the midrange portion of the signal too much?

I'm curoius to know if any improvements have been made in these areas since the shooting of the Peter Jackson's short and/or if a bit more grading would have made a difference to these problems as I understand they didn't spend very much time on that original grade.

Has any else felt this way about the images is or is just me? Am I being too critical because there's no gate weave or grain to distract me anymore?

[Gavin Greenwalt]

Are you sure you saw 4k footage. I'm pretty sure they only screened 2k 4:2:2 footage yesterday.

[ChristopherKenworthy]

I saw the 1080P footage at SMPTE a few days earlier, and was quite alarmed that the clouds looked really quite badly clipped. I was trying to look for faults, however, rather than watch the story. I remind myself that these were alpha cameras, tested in a rush, with almost no features enabled and a board that's now been replaced. Given what Ted was hinting at, with regards to a latitude that exceeds some film stocks, I don't think it will be an issue with the final release.

[Álex Montoya]

At the time of the release of the short film it was stated that the latitude when shooting was only about 8 stops and that it has been since largely improved.

[ChristopherKenworthy]

At the time of the release of the short film it was stated that the latitude when shooting was only about 8 stops and that it has been since largely improved.

There you go. Couldn't remember that, but knew we were in that sort of arena. I know I wasn't hallucinating when I heard Ted on Wednesday talk about getting towards 16 stops. So, that's twice as good as Crossing The Line. Gulp.

[Álex Montoya]

16 stops? With the existing sensor? That would certainly change drastically the way of shooting...

[ChristopherKenworthy]

16 stops? With the existing sensor? That would certainly change drastically the way of shooting...

Ted said a few things on and off the record. Bear in mind that I had been awake a solid 29 hours when I met Ted, so my perceptions could be distorted. I stormed up to him before he'd had time to set the camera up and threw questions at him. He's a lovely guy, and I got lots of very, very interesting answers. But the thing about 15-16 stops came up at the big presentation at the Apple stand later that day, and I think he repeated it elsewhere during the week. He said that latitude is difficult to measure, and some people say film has a latitude of around 15-16 stops, and that Red was hoping to get there before too long. (This camera, this sensor, later update? Who knows?) Later, he used the word "eclipse"; I remember it clearly. He said that Red was hoping to have enough latitude to "eclipse" some popular film stocks. I got the impression he was talking about this sensor and the camera we will all have in our hands. I am still doing a double-take on this information, wondering if I've dreamed it all, but I think I've seen it elsewhere in the forum, where people are talking about Australian SMPTE. If 15-16 stops are possible, with this version of the camera (even if it requires a firmware upgrade) it is major news. I should also add, that he stressed over and again that the initial releases will not be fully enabled (as we all know) so if this 15-16 stops is ever true, it may be some time away. Still, it's a potent thought.

[Desert Rune]

I know I wasn't hallucinating when I heard Ted on Wednesday talk about getting towards 16 stops. So, that's twice as good as Crossing The Line. Gulp.

You mean 256 times as good. :angry01:


:w00t:

[Graeme Nattress]

The colour in the clouds was added in grading, I do believe, to make things look more like sunrise. The entire visual look of the piece, what the skin tone was like etc. etc. was down to grading choices by PJ's team. We just gave them the raw pixels to work with, so to speak.

It's hard to be criticial in the way you are being of a camera, when you cannot see the raw footage as shot and were not at the grading session with the director. I've seen the raw footage, obviously, but I was not at the grading session and do not know the directors intent.

Talk of dynamic range is just that - talk. We've made real improvements in how RED One is working, as can be seen by Jim's ISO tests that he posted recently, however any talk of numbers or where the numbers might go in the future is pure speculation.

Graeme

[ChristopherKenworthy]

The colour in the clouds was added in grading, I do believe, to make things look more like sunrise. The entire visual look of the piece, what the skin tone was like etc. etc. was down to grading choices by PJ's team. We just gave them the raw pixels to work with, so to speak.

It's hard to be criticial in the way you are being of a camera, when you cannot see the raw footage as shot and were not at the grading session with the director. I've seen the raw footage, obviously, but I was not at the grading session and do not know the directors intent.

Talk of dynamic range is just that - talk. We've made real improvements in how RED One is working, as can be seen by Jim's ISO tests that he posted recently, however any talk of numbers or where the numbers might go in the future is pure speculation.

Graeme


Completely understood. Talk is talk, and grading is grading. And gut feeling is gut feeling. Personally, I have no worries.

It reminds me of a screening of Flyboys footage in Palm Springs. The director had chosen to blow the sky out completely in some shots, but no matter how the Genesis people pointed out that this was a director's choice, the audience kept muttering for days about the "blown out sky".

[Michael Mann]

I know I wasn't hallucinating when I heard Ted on Wednesday talk about getting towards 16 stops.
16 stops with 12 bits? Isn't 12 stops the limit? Thanks for clarification in advance.

[Stephen Williams]

16 stops? With the existing sensor? That would certainly change drastically the way of shooting...

Hi,

Twice as good would be 1 more stop!

Stephen

[david farland]

I heard 'Crossing the Line' at NAB was output in ITU.709 for Sony projector.

Footage shown at AFTRS in Sydney was ProRes 2K 4:2:2.

Cheers,

[Cail Young]

Footage shown at AFTRS in Sydney was ProRes 2K 4:2:2.

That's disappointing, that's exactly what I saw at the conference screening, then (not the public one, the one in the auditorium on the oddly shaped screen).

From what I saw at Future Reality's stand a lot of the shots were pushed very hard in the grade (that, or the default one-light in REDCINE is really dark).

[John V]

ahhhh remember... 2days prototype camera..think about it guys..not finished product. It would be like complaining because the car does not move because it has no engine.

[John V]

Completely agree. Wait till its done and knowing Jim it will never be complete even after you get it..upgrades...love the foresight.

[Stephen Gentle]

The Onfilm article (http://www.onfilm.co.nz/editable/DigiFeature.pdf) that was recently posted online shows how Crossing the Line was filmed.

The only camera controls they had were 'go' and 'stop'. And with the new boards that were done in the engineering delay, it seems that the dynamic range has increased a few stops!

[J. Bernard Vallon]

16 stops with 12 bits? Isn't 12 stops the limit? Thanks for clarification in advance.

Encoding into 12 bits limits DR to 12 stops, but considering how difficult it must be to reduce noise low enough to get 15-16 stops, switching redcode raw from a 12 bit codec to a 16 bit codec is probably relatively simple (meaning it can be done and, presumably, graeme and co know how to do it)

[Gavin Greenwalt]

You can compress 32 stops into 8 bits... but you probably wouldn't want to (very very flat image). 12 bits for 12 stops is just the limit at which point you are no longer recording useful information and you're just capturing noise. And since you don't want to risk throwing away any information that you don't have to--the absolute mathematical-maximum information possible option is probably the best.

Film has more than 12 stops of latitude and to date it's been almost exclusively recorded to 10bits.

[Nick Shaw]

Film has more than 12 stops of latitude and to date it's been almost exclusively recorded to 10bits.

But that's 10 bit log. RED's 12 bits are, as I understand it, linear.

[Brook Willard]

You can do 10-bit log if you wanted... but it doesn't really make sense. A sensor is inherently linear [analog].

[Iron Possum]

Don't you love these guys who come outta no where, and on their very first post they make some sort of criticism about Red, and then that's it. . . we hardly ever hear from them again.
What do they call it in New Zealand and Australia? "Tall Poppy Syndrome"?



Sorry about not chiming in again sooner but I've been flat chat all weekend. I just wanted to send my query off about the clipping and skin texture before I forgot about it.

It's not a tall poppy thing at all, but when you hear nothing but good things about a product and don't see or hear any criticism whatsoever then that's not such a good thing either. I believe the concerns that I raised were valid observations and since I didn't get to ask about them during the screening I should be able to go ahead and raise them here.


It's hard to be criticial in the way you are being of a camera, when you cannot see the raw footage as shot and were not at the grading session with the director. I've seen the raw footage, obviously, but I was not at the grading session and do not know the directors intent.

Graeme

Thanks Graeme, it's good to know that latitude has been improved since then and that the skin texture issues were just down to grading choices. I'm sure you know the look I'm talking about and if you say it wasn't there in the raw images then that's good enough for me. Coming from a predominantly video background I am constantly frustrated by that "plastic skin" look and was, as you may understand a little disappointed when I noticed it at the screening, especially since the camera has been hyped so much.

None the less it does seem like an amazing camera for anything near it's price point. I very much look forward to seeing further test results in real world situations over the next couple of months.

[Cail Young]

The Onfilm article mentions that the film was graded in 709 linear space, which may explain the weirdness of the sky. They are apparently now regrading in log, which may provide better images (whether we'll ever see them is a different matter).

[Michael Mann]

You can compress 32 stops into 8 bits... . Understood! Thanks.

[Graeme Nattress]

Cali - REC709 isn't linear, it's gamma encoded. Quite frankly, "what space you grade in" doesn't make a bucket load of sense to me, as any high end app will take the data directly into float anyway. How the data is represented is not the issue, but the accuracy of that representation is, and, how the grading mathematics work.

In all the raw files I've crunch, I've not noticed anything that I'd say was untoward about skin texture. To me, the camera footage always seems to look like a high end DSLR image (which is an image I like) that has not had much in the way of tweaking done to it - which is what you want going into a grading session.

[Cail Young]

Cali - REC709 isn't linear, it's gamma encoded. Quite frankly, "what space you grade in" doesn't make a bucket load of sense to me, as any high end app will take the data directly into float anyway. How the data is represented is not the issue, but the accuracy of that representation is, and, how the grading mathematics work.

In all the raw files I've crunch, I've not noticed anything that I'd say was untoward about skin texture. To me, the camera footage always seems to look like a high end DSLR image (which is an image I like) that has not had much in the way of tweaking done to it - which is what you want going into a grading session.

I was just quoting the article - I'm still wrapping my head around the concept of colour spaces in general...

[Mike Seymour]

I was involved with helping Ted and certainly arranging two of the three screening setups.

But in all three locations, the SMPTE conference, Apple's booth and the ACS night the footage was 2K, but at the conference and at ACS it was played on a Sony 4K projector. We worked hard to try and get the projection right - at the ACS in particular we spent a lot of time adjusting things and checking sound etc, to give the audience the best chance to view to the material. Ted offered to project it 4K but there was problems with the availability of the servers to play back the 4K to the projector.

Of course it would have ben ideal to project 4K, and we tried for 2 solid weeks to work it out, but Sydney does just not have many (any) 4K servers and both ACS and SMPTE are non-profits so we were all working to do the best we could. Once we understood how impossible it was to project 4K DCP level, we still used the very best 4K projectors we could thanks to SONY and I believe that the projection did the film proud - based on having seen it 4K at NAB on a smaller screen.

Actually I found it very interesting to see the 2K ProRes version, as it was stunning how great it looked given the reduced data rate. In a sense I felt like the ProRes confirmed RED as an everyday option and not just a 4k option. If that makes sense.

I would just quote Ted given Crossing the Line was shot on prototypes "this is the worse the images will ever be" meaning any footage shot from now on will benefit from the big improvements made since March/April.

Just a thanks to David Wakely and all the team at the ACS who worked so hard to put on the event in Sydney and the SMPTE for the conference, and Lucas Wilson from Assimilate for showing Scratch and REDCINE.

Mike
fxphd.com

[Scott Webster]

Great work Mike and nice of Sony to assist in making the event a success.

Will ProRes still be relevant once Redcode is a codec choice in FCP?
Do Redcode and ProRes recognise each other? is it simple as plugging in a Red Drive via FW and dropping the files into FCP and selecting ProRes?

[Mike Seymour]

There are two paths you can choose from - Redcode or ProRes (or anything else I guess) If you were finishing in FCP for a 720P master - you may convert to ProRes and just finish out that way, but you could also run a timeline that is Redcode - of course I work for neither Apple or RED and this is not released yet but it is clear that you can work either way from the discussions I have had. Thats the short answer - the long answer will depend on what you aim is - what your plan is for the material (final resolution etc) - plus your budget and thus system capacity.
As much as it is fun to think about it, very few people will shoot 4K, edit 4K and project 4k at least for a while, so while it might be cool to want to maintain 4K - in reality it will be so much easier to sample to 4K (and then be able to go back to it if needed for a zoom etc) - but actually not finish a 4K master everytime.
IMHO

Mike

[M Most]

Cali - REC709 isn't linear, it's gamma encoded. Quite frankly, "what space you grade in" doesn't make a bucket load of sense to me, as any high end app will take the data directly into float anyway. How the data is represented is not the issue, but the accuracy of that representation is, and, how the grading mathematics work.


I would say that it can be important, but it depends on what your intended primary delivery format is. If it's a digital cinema or other digital delivery format, a linear (or, in most cases, video gamma encoded) format is fine. If your primary deliverable is a film print (and for the near term, it will be in many cases), that changes somewhat in that you really want to be seeing a reasonable representation of your film output - which, in most cases, is going to be best represented by a log encoded image projected through a film print preview LUT. How you get to that, however, is another story.

[Mathew Mackereth]

the images we saw at AFTRS on friday on the 4K projector were definitely 2k in the proRes format playing of a laptop...

[Iron Possum]

Okay, I remember that the AFTRS screening on Friday was 2k Pro Res going through a 4k Projector. I'm sure Pro Res wasn't responsible for what I saw, but as Graeme has mentioned these were early prototype camera's and have come along way since then, which is good news. It's reassuring to here that the images are not unlike those from a good DSLR. I love the images I get from my camera, so I feel a lot better knowing that.

I'd also like to thank Mike and David for organising the demo last week. We've had a couple of good things from you guys in the last couple of weeks, keep it up. Speaking of which, when do we get to see the results from the Green screen tests we did the other week with that other camera?

[Gavin Greenwalt]

The Onfilm article mentions that the film was graded in 709 linear space, which may explain the weirdness of the sky. They are apparently now regrading in log, which may provide better images (whether we'll ever see them is a different matter).

Oh no! I knew that quote was going to cause a lot of problems.

I have no idea what they were talking about there. And I've heard the quote thrown around a great deal since then... and still have no idea what that means.

All the software I know of grades in linear space behind the scenes regardless. I don't know what grading in "log" will actually deliver. And I'm very suspicious it's just grader superstition. However I would be very curious to hear otherwise.