Pretty easy to understand...

[Cory Mitchell]

What bothers me more than anything is how other companies hold back their technology so they can put it in next years update. I LOVE that RED does not do that. Red puts in the best they have at that time, and when they realized that they were going to continue to progress they changed and made a modular system. Canon and the C300 is a great example. They've already announced the C500, even though the C300 still new. They had the technology, but held on to it when they realized (from my point of view) that their C300 wasn't good enough (it was not future proof at all).
I remember watching Panasonic, Canon, JVC and Sony for their new cameras and always being disappointed when they only now added stereo sound, or added 1080p that wasn't really 1080p, or 720p that was two 640x720 sensors so that each half of the picture was slightly different. And the whole 1080i thing. Ugh. I understand that they had their markets but their upgrades were only ever incremental, with technology they could have put in the previous model. I remember seeing real 1080p pro cameras for $100k+.

And then Red came along. I know it's not all about resolution, but the point is that they aren't holding back cards so they can incrementally update over the next 5 years to technology they have today.

Thank you Jim and Co, for really pushing the envelope instead of trickling down mediocre technology in incremental upgrades. Thank you.

 

[Carlos Florez]

Thanks Patrick!!!

WoW WoW!! Carlos......Great Short Film :001_tt1:

 

[Ricardo Moreno]

Not exactly sure what you just said but I would certainly debate C300 "looks and feels nicer". Really?

Jim

I thing I get it. I guess he means the monitor on location. I used to get it with the red one at the beginning from clients with purposely flat R3D files in RAW preview. Clients didn't feel very good about seeing the raw as it is, raw, but with this is not an issue an this is not an issuewanymore. Clients are used to and RED has improved the Raw preview much better.

I thing I get it, I guess he means that some video guys might prefer the other guys, because they don't understand the RED workflow. They are used to shot and edit. The other guys could look nicer in the monitor form the getto, but that is it, they are so limited no where else to go, what you see is what you get, they forget that there is a very important part of the raw workflow in post with zillion of choices that no else has but RED.

 

[Jon MIchael Puntervold]

Of course, all of this is subjective - many here would prefer the "natural look and feeling" of the Epic over Alexa - and in any case influenced much more by the colorist (addendum: and cinematographer, director and production designer) than the camera.

I don't agree. The camera itself has a look, and most cinematographers seems to admire the look of Alexa, or at least there is a notion that the camera is more forgiving, atleast compared to the Red One... But I'm sure Dragon will improve digital cinematography even more...

 

[Brad Allen]

I dunno... I know a lot of people who shot broadcast promos on RED RAW... then said "WTF?! It'll take this long to transcode my files?!"... and then chose Alexa LogC ProRes for the next shoot.

It's so bizarre for me to hear comments like this when I'm coming from an Adobe workflow. The idea of needing to transcode anything before final render is just foreign to me. The most frustrating aspect for me is that people would write off an entire camera system because the software they use hasn't caught up to what Red has been doing for the last 5 years.

 

[Erich Ocean]

It's so bizarre for me to hear comments like this when I'm coming from an Adobe workflow. The idea of needing to transcode anything before final render is just foreign to me. The most frustrating aspect for me is that people would write off an entire camera system because the software they use hasn't caught up to what Red has been doing for the last 5 years.

You need an incredibly fast machine for your comments to make any kind of sense.

With my 3.4 GHz Intel Core i7, I can't even play back 30fps 3K material at quarter rez without hiccups in Resolve. RedCine-X Pro I can't even use for playback, it buffers like one second, then plays that, then fills up the next buffer for a few seconds, then plays that, etc.

Adobe Premiere is like Resolve, quarter rez almost plays back in real time.

1/8 rez does play back, but looks like crap.

Like Bruce says, you need to either spend $5K on a RedRocket, or you need to spend $5K on a computer that still can't play back an R3D in full quality in real time.

 

[William Robinette]

You need an incredibly fast machine for your comments to make any kind of sense.

With my 3.4 GHz Intel Core i7, I can't even play back 30fps 3K material at quarter rez without hiccups in Resolve. RedCine-X Pro I can't even use for playback, it buffers like one second, then plays that, then fills up the next buffer for a few seconds, then plays that, etc.

Adobe Premiere is like Resolve, quarter rez almost plays back in real time.

1/8 rez does play back, but looks like crap.

Like Bruce says, you need to either spend $5K on a RedRocket, or you need to spend $5K on a computer that still can't play back an R3D in full quality in real time.

Interesting Erich,

I have a basic 15" 2.2 i7 MacBook that plays back in REDCINE-X Pro and Premiere Pro at 1/4 res with no hiccups unless I'm using slow drives via USB2 or something. FW800 still hiccups sometimes but Thunderbolt (and I would assume eSATA as well) works great. Not full res, but definitely 1/4 with no problems. Perhaps something else is going on? I'll agree it's nothing you want to do mission critical clients over your back type stuff, but for desktop edits?

But back to topic, I'm glad I'm filming everything in 4k now. And that you Jim, for making it in good old CA. Cheers!

 

[M Most]

Thanks Michael. Yes they could do that... but time is running out!

My other thought is if RED really wants people to work with RAW R3Ds all the way... they should just dump the whole hardware decode box idea - and design their next version of REDCODE so that it can be decoded well by GPUs.

If something inherent to REDCODE means that it can't be decoded by GPUs or CPUs... then maybe they need to change REDCODE?

According to what I've been told - and what has been posted here, although I can't recall exactly when - the issue is not with the debayering/frame reconstruction, it is purely with the decompression. Wavelet compression is rather compute intensive, and in most cases does require hardware assistance to achieve real time encoding and decoding performance. This is true of essentially all JPEG2000 implementations. The only real exceptions I can think of to this are Cineform, which is somehow able to achieve real time performance on commodity hardware, and the decoding engine that Iridas developed a few years ago that was heavily GPU enabled (and as I recall only provided real time playback at 2K or lower resolution). But every flavor of J2K that I know of - including J2K compressed frame sequences recorded on a Codex device - requires hardware assistance for full resolution playback in real time. There is a CUDA enabled version of J2K, but I don't believe it achieves real time playback at the kind of resolutions that we're talking about here, either. This is probably one reason why the manufacturers that you mentioned (Sony, Arri, Canon) have chosen not to use J2K compression in their RAW recordings. Arri records RAW uncompressed (well, actually they supply uncompressed RAW to third party recorders that are free to use whatever they want), Sony is using compression that is based on MPEG4 simple studio profile, and Canon is deferring to third parties to come up with RAW recording/compression solutions.

Jim has repeatedly pointed out that this stuff isn't easy. That's just as true for recording and compression schemes as it is for the camera itself.

 

[Brad Allen]

You need an incredibly fast machine for your comments to make any kind of sense.

This just isn't true - I'm sorry that your system isn't working as well as you like - but my machine isn't particularly fast or expensive. I too am running a core i7 - quad core - 3.06ghz. Cost about $400 (for the CPU) two years ago. Playing back 1/4 res in Premiere is no problem at all. To be honest, it runs as smoothly as any sort of proxy would.

Perhaps Resolve is a different story - but usually a grading system is specced differently from all the NLE's in an editing house. Additionally, Bruce's point about transcoding R3D's didn't apply to final grading I didn't think. Any colorist choosing to grade with RGB baked proxies as oppossed to RAW R3D's probably doesn't understand just how much they are limiting themselves when it comes to image manipulation (to put it kindly).

RedCine-X isn't quite as well optimized for multithreaded operations I've found - but even then I don't know that it's much of a problem as I don't think most people are using RedCine-X for continuous playback of R3D's - it's not built as a NLE that needs to be able to give real time playback. Having said that, unless I'm using post-processing effects in RedCine, I find that 1/4 resolution is able to be played back very quickly on my system.

Maybe I'm missing something but R3D's have been extremely easy to work with on my end.

 

[Blair S. Paulsen]

(snip) My other thought is if RED really wants people to work with RAW R3Ds all the way... they should just dump the whole hardware decode box idea - and design their next version of REDCODE so that it can be decoded well by GPUs. (snip) If 20 megapixel 3:1 compressed F65 RAW can be played back by the free version of DaVinci... what is Sony doing right that RED is doing wrong?

Bruce allen
www.boacinema.com

Let's revisit this issue in about a month. The prognosticator in me expects the landscape to change from at least two different angles very soon. I believe that RED is very interested in making the world of r3d post much better and will have some enabling products on display at NAB.

IMHO RED was waiting for several things to happen before they pushed beyond the Red Rocket:

PCIe 3 (on both the PC side and the next gen Mac bombers)
Newer CPU/mobos, IO and GPUs
Enough people using r3d media to interest key partners

If I had to guess I would predict a reworking of RedRay so among other features it can act as a black box that plugs directly into a PCIe 3 slot in addition to lots of other IO options. Why?

Manufacturing costs - particularly the chip sets that decode/scale/encode, etc. The ability to offer a unit that does just about everything internally to avoid bandwidth/signaling issues, easier to control IP, etc. They might go modular to ease future hardware upgrades but I would predict proprietary interconnects.

I think the really interesting question is price point. For Bruce (and those who share his issues) its hard to sell an expensive box that only does one thing. Soooo, do you sell it cheap to flood the market - make your $ on volume and spark the growth of the RED world? Do you make it capable of zipping trough other codecs and feeding out 4K from other file types so it becomes a feeder for 4K displays no matter what codec the footage came in as?

The next couple of days should be interesting... ,-)

Cheers - #19

 

[Scott Summers]

I'm pretty positive he's just saying it's hard to argue with people's subjective thoughts or feelings. If they say "brocolli tastes nice, and you say "no it doesn't", you can probably argue till you're blue in the face but they'll still think it tastes nice.

He's saying if someone thinks the C300 "looks or feels" a certain way, it's hard to combat such subjectiveness. Obviously we can throw as much technical jargon at 'Mr Feel' as we want, but he's still going to say, "yeah but I...I just like the FEEL of it"... and that's where you throw your head into a brick wall.

Not exactly sure what you just said but I would certainly debate C300 "looks and feels nicer". Really?

Jim

 

[M Most]

Perhaps Resolve is a different story - but usually a grading system is specced differently from all the NLE's in an editing house. Additionally, Bruce's point about transcoding R3D's didn't apply to final grading I didn't think. Any colorist choosing to grade with RGB baked proxies as oppossed to RAW R3D's probably doesn't understand just how much they are limiting themselves when it comes to image manipulation (to put it kindly).

Or they want better playback performance than they're likely to get using R3D files directly, especially when multiple streams are required. Or they're working on projects that have a significant number of visual effects shots and/or shots done with cameras other than Red and don't want to deal with multiple formats and the resulting inconsistencies. Or they're working on a show in which the editors used a lot of editorial effects that need to be pre-rendered. Or they've realized that even though it sounds good to have the flexibility, the fact is that on about 95% of the material they grade, they rarely if ever change the settings anyway and instead use a combination of Redcolor2 (or 3 now) and Redlogfilm, which essentially gives them the same starting point they would have if they were using the R3D files directly with the same settings - but with much leaner use of computer resources.

I used to believe as you do. And although I sometimes will still do a show directly from the R3D files, I've done an awful lot of them now without that. And I've never really felt particularly limited, at least not since Redlogfilm showed up.

 

[Adam Beck]

Companies that actually deliver the future are better than those that pretend the past "is good enough".

Jim

Wow......this has so much truth behind it.

 

[Zakaree Sandberg]

you win.. don't worry

 

[Brad Allen]

Or they want better playback performance than they're likely to get using R3D files directly, especially when multiple streams are required. Or they're working on projects that have a significant number of visual effects shots and/or shots done with cameras other than Red and don't want to deal with multiple formats and the resulting inconsistencies. Or they're working on a show in which the editors used a lot of editorial effects that need to be pre-rendered. Or they've realized that even though it sounds good to have the flexibility, the fact is that on about 95% of the material they grade, they rarely if ever change the settings anyway and instead use a combination of Redcolor2 (or 3 now) and Redlogfilm, which essentially gives them the same starting point they would have if they were using the R3D files directly with the same settings - but with much leaner use of computer resources.

I used to believe as you do. And although I sometimes will still do a show directly from the R3D files, I've done an awful lot of them now without that. And I've never really felt particularly limited, at least not since Redlogfilm showed up.

There's some excellent points here Mike. There are of course exceptions to every rule - many of which you have pointed out. I know you're the man when it comes to knowing what actually happens in many post production facilities, though I'm not sure that many of these problems are too big to be overcome.

For those projects that are exclusively Red material, are most grading systems you're working on struggling with the processing demands of R3D's? Even when only previewing @ 2k? I know you're experience here is greater than mine so I'm happy to default to your answer. From my experience, I thought most grading systems were pretty well equipped to do R3D's online these days, but again, you'll know better what most facilities setup's look like.

Regarding FX shots, it seems an odd reason to avoid R3D's because of inconsistencies. Are we talking inconsistencies in the way that software deals with R3D's vs. Prores, DPX, OpenEXR etc. Or are we talking inconsistencies in how well the footage can be graded to match?

Multiple formats, I can understand - but in that case it's not as though R3D's are at fault. Every camera would be converted to the same format, resolution, bit rate etc. I know you're not highlighting R3D's as the problem here Mike and were just pointing out exceptions to the rule - just wanted to make sure others didn't confuse Red files as the issue behind this point.

Effects for editorial content I'm actually a little confused on, what sort of effects are we talking about? Plugins (like the sapphire effects), transitions, lower thirds or something else entirely? Really interested in this one actually as I just can't think of a whole lot of times editorial is doing effects work that is rendered out. What % of shots do you have coming into a grade or finishing session that have been placed there from editorial?

I take your point regarding the fantastic results you can get out of a RedLogFilm transcode, but even if the results are acceptable - surely they aren't as ideal as having access all the RAW data when grading, which I guess is why you still find your self grading with R3Ds from time to time. That and the fact that you don't have to transcode all of your footage to start with? I guess what it comes down for me is that if the horsepower is there for an online finish in raw, then it doesn't seem to make sense in many cases to transcode everything to start with - especially if the NLE's for editorial can read the R3D's without hiccups.

 

[Gunleik Groven]

The big issue of R3Ds is not the debayering but the encryption, and that is really an element that makes REDRAW NOT a good choice for a lot of applications, despite its superior image offerings...

 

[Patrick Grossien]

The big issue of R3Ds is not the debayering but the encryption, and that is really an element that makes REDRAW MOT a good choice for a lot of applications, despite its superior image offerings...

REDRAW MOT?
What are you referring to Gunleik?

Thanks
Patrick

 

[Mathieu Ghekiere]

I think he means, that makes REDRAW 'NOT' a good choice...

 

[M Most]

For those projects that are exclusively Red material...

I've never worked on any show that was exclusively Red material. I don't think I've ever even encountered one.

...are most grading systems you're working on struggling with the processing demands of R3D's? Even when only previewing @ 2k? I know you're experience here is greater than mine so I'm happy to default to your answer. From my experience, I thought most grading systems were pretty well equipped to do R3D's online these days.....

The logic of grading directly from R3D files makes sense only if you happen to be conforming on the same box you're grading on. In most facilities - especially in television - this is rarely the case. Shows are usually conformed on Avid equipment, be that Symphony or DS, primarily because that's the best way to deal with the things editors do today. Keyframed speed changes, Sapphire plug in effects, multiple panels on the screen - these are all things we now see in abundance in just about every show we do. The only sensible way to conform such shows is to do it on a system that doesn't require deconstructing and recreating all of these things, because to do that is, quite frankly, much too time consuming. And that's time that the clients refuse to pay for. The most common way of working is to "conform" on Avid, and use the output of that on the color systems, along with a simple EDL to cut it up. Now, I should also point out that in television, we're almost always dealing with the Alexa as the "A" camera these days, so this approach really doesn't have a downside. But even on a Red show, the same requirements apply.

Regarding FX shots, it seems an odd reason to avoid R3D's because of inconsistencies. Are we talking inconsistencies in the way that software deals with R3D's vs. Prores, DPX, OpenEXR etc. Or are we talking inconsistencies in how well the footage can be graded to match?

I don't know how many VFX vendors you've dealt with, but we've found that just about every VFX vendor, from the most established to the newest, seems to come up with their own way of dealing with the files you give them, and what you get back is almost never what you sent, even when you specify exactly what you want and/or need. Since these things are so unpredictable - regardless of how many phone calls or meetings you have to go over it - you're better off not depending on specific results, and not expecting seamless matches.

Multiple formats, I can understand - but in that case it's not as though R3D's are at fault. Every camera would be converted to the same format, resolution, bit rate etc. I know you're not highlighting R3D's as the problem here Mike and were just pointing out exceptions to the rule - just wanted to make sure others didn't confuse Red files as the issue behind this point.

R3D is not really any more problematic in itself, but it does allow the user to reinterpret its results by changing a few settings. Therein lies the heart of what I just referred to. Some vendors use the camera settings, whatever they happen to be. Some insist on using an "in house" recipe of, say, Redcolor and Redgamma. Others use Redcolor 2 and Redgamma 2. Others use, well, you get the picture. And it doesn't seem to matter what you tell them to use, they all have what "they like" whether it matches what you're using or not. RedlogFilm should have been the great equalizer, but it hasn't worked out that way because far too many people (and vendors) have little to no understanding of a proper processing pipeline for log images, so they either treat them like video gamma images (crushing lots of detail and creating an odd greyscale in the process), or invent their own curve corrections that have no standardization and no relation to anything other than what "they like."

That's SOME of what I meant by "inconsistencies...."

Effects for editorial content I'm actually a little confused on, what sort of effects are we talking about? Plugins (like the sapphire effects), transitions, lower thirds or something else entirely? Really interested in this one actually as I just can't think of a whole lot of times editorial is doing effects work that is rendered out. What % of shots do you have coming into a grade or finishing session that have been placed there from editorial?

Split screens, multipanel shots, speed ramps (often with keyframes), odd color effects, plug in effects - these are some things we see. In abundance.

I take your point regarding the fantastic results you can get out of a RedLogFilm transcode, but even if the results are acceptable - surely they aren't as ideal as having access all the RAW data when grading, which I guess is why you still find your self grading with R3Ds from time to time. That and the fact that you don't have to transcode all of your footage to start with? I guess what it comes down for me is that if the horsepower is there for an online finish in raw, then it doesn't seem to make sense in many cases to transcode everything to start with - especially if the NLE's for editorial can read the R3D's without hiccups.

The times I use the original files are usually when it's a job I'm doing on my own and can make my own choices. In my "day job" none of this is the colorist's choice. But even if it was, I would probably opt for what we do if it is a television project, because the combination of turnaround requirements and client expectations (they expect to watch a fully conformed and finished show in real time, without playback hiccups of any sort, in the color room - in fact, we play out to the master tape delivery element from the color systems...) make it the best approach. On a "side" job, especially one that is a DI, I might go with original files, or I might not. It really depends on the material, the amount of non-original material that will be integrated, and the particular client's expectations as to what they want to see and in what environment. Every DI I do seems to be a bit different and really demands a bit of rethinking as to a proper finishing path.

 

[Gunleik Groven]

REDRAW MOT?
What are you referring to Gunleik?

Thanks
Patrick

Ah... Typo...

as in "not"

As M Most is refering to, a "RAW" grade is not allways the best solution, and RLF has made life a lot easier for non-RAW onlines...
We do RAW now, mostly because of diskspace, not quality, really. But the stations needed for doing that are rather costly, actually...

And it is a bit irritating in that respect - to know that a lot of the computing overhead is going into de-encryption and not image processing.

How can I say that?
Well...
We kinda had good speeds on the redRAWs before the encryption was introduced. I do understand and respect why it was introduced, but it is an achilles heel of the system for a lot of applications.