DRAGON - ITS THE REAL DEAL !!!!

[David Battistella]

So, by that rationale, those who have concerns about perceived weaknesses of the sensor should wait until RED has had time to address those issues, but those who want to celebrate and proclaim it as the wonder of wonders can do so right away? Hummmm....

No. That isn't what I was saying. I think the discussion is good and healthy but also has to be looked upon as a whole. I don't have to defend Mark, RED or Jim or anyone, they are big boys, they can take care of themselves. I just know from years of looking at RED images, knowing what my camera can do and where it breaks down (since day one it care needs to be taken to not try to clip highlights (but I would say that is a digital thing, although others will say that ARRI cleverly resolved this in their OLFP), but there are other compromises to consider there.

I'm not even asking anyone to give RED or Mark a break.

What i see in those images is great potential and an advancement from what I have.

Others, well. They just see highlight clipping. There are compromises with any camera you pick up. Literally, this is an endless conversation. I see what I see, you see what you see.

I agree that a lot of people will get on the naysayers here (often to a fault) and feel the need to defend (not sure why) but I am seeing this as a constructive conversation to which I am contributing thoughts and observations.

Like you. And I respect that.

David

 

[Elsie N]

Quote of the year right there, that about sums it all up. Thanks.

Yeah, I would imagine any major weaknesses have been identified and are being addressed via firmware and Color Science tweaks. It's my considered opinion this company has changed from the "let the customer help build it" one they once were and are now this well oiled machine that understands what being an industry leader, rather than just an innovator, entails.

 

[Martin Stevens]

14+ stops of DR is quite enough!!!

There is no sensor that has 20+ stops of DR on the market.

RED should focus on colour science and after 6 years finally to fix it!!!!!

Sanjin.....there is nothing wrong with the Dragon color science that cannot be altered or fixed via grading.

 

[Andy Roberts]

If this were meant to be a celebratory thread only, maybe a stress test was not the best route to go for Dragon's coming out party.

Red didn't ask Mark to do a stress test. That was his choice.

All I'm suggesting is that critiquing a compressed Vimeo video just seems odd to me.

As for me, I tip my hat to Red and Crew, then Mark Toia, and carry on.... with my $9,500 patiently waiting to be exchanged.

 

[David Battistella]

Mark is not the only person who is testing. It's just the first public stuff presented here.

David

Darius Wolski, Inez and Vinoodh, Chris Probst, Peter Jackson + Andrew Lesnie, , Michael Bay and Amir and Peter Collister all have wielded the Dragon. Toia is up to bat next.

Most of these guys priority with the Dragon isn't to take "marketing" images.. they are shooting tests or footage for their current or next films.

But don't worry... im sure a few people like Toia are gonna have a hard time keeping footage to themselves

 

[Carey Lee Coffey]

Let me clarify:

Light is necessary to capture an image. Obviously. What's overrated is the importance beauty lighting, or super slick Hollywood-style (artificial) lighting plays in creating a filmic image. Of couse you need light and you need to manipulate light and shadow in order to get the image you are after, but the lighting should not make everything look so beautiful, so perfect, that it looks fake. Unless, of course, that's the look you are going for. Personally, I detest the look of a lot of modern films because they are so perfectly lit that it pulls me out of the story. There are exceptions. It's totally subjective. Dragon enables you to shoot with less light. That's the future.

 

[Phil Holland]

14+ stops of DR is quite enough!!!

There is no sensor that has 20+ stops of DR on the market.

RED should focus on colour science and after 6 years finally to fix it!!!!!

Sanjin.....there is nothing wrong with the Dragon color science that cannot be altered or fixed via grading.

I agree with Martin here, but I'll also add this into the mix.

The increased dynamic range, usable latitude, and lower noise floor within the REDCODE is going to provide a great deal of flexibility in pushing around the tonality and chroma within your image.

This is important to me and one of the real selling points behind having this much range.

So to counter your statement there Sanjin Red is focusing on their color science a lot right now. Having this amount of latitude is part of that.

 

[Timur Civan]

I think what people are missing from this is that when actually lighting, we will rarely create a scene with 15+ stops of latitude. What we will benefit from is a beautifully rendered middle 12 stops with little fussiness or distortion. This is why the Alexa has "amazing" feel, is that high end DP's who light with it have the luxury of letting a highlight go a little more, but still have it behave.

The dragon takes that to a new level in the red ecosystem. The current MX is about 12.5-13 stops to clip with a very well behaved 10 stops in the middle. That's a lot actually. The Alexa is a well behaved 12 with a mushy top and bottom. Yea it rolls off nice, but that rolling top stop and a half is not actually rendering anything too substantial. Arri just programmed the curve to burn to white in a nice way.

Based on what I've seen, from this test, and my own experience shooting races, beaches and Forrest locations, yes the dragon burns the hard edge of its clip a bit harsh, but that's easily remedied in a simple grade, and the edges can be softened with the click of a button or with a 1/8th black magic. BT you have to remeber, the only burned edges we've seen are the direct reflections of the fucking sun. Or at least a low angle reflection of the direct sun on water, Film won't hold that much detail. I know, I shoot tons of stills on film. Just to see the difference. This is 400 speed Portra. With a direct sun backlight, aaaand a flare lifting the shadows, and its still blowing out. The dragon would hold this easily.... Would the dragon have as much character? Maybe not. This is chemicals interacting with photos....

That said I don't understand why folks seem disappointed..... I will be freaking ecstatic to shoot dragon.

 

[Andy White]

I think what people are missing from this is that when actually lighting, we will rarely create a scene with 15+ stops of latitude. What we will benefit from is a beautifully rendered middle 12 stops with little fussiness or distortion. This is why the Alexa has "amazing" feel, is that high end DP's who light with it have the luxury of letting a highlight go a little more, but still have it behave.

That said I don't understand why folks seem disappointed..... I will be freaking ecstatic to shoot dragon.

I've held off posting until now (on this particular section of the debate... but you're right.

Problem is, people are comparing Mark's 'stress' test (with the wildest of vari-ND and shooting where he wanted to push things) against perfectly lit and 'controlled' shoots.

The highlight rolloff on the Alexa is better... well, please (please) show me a comparable 'stress' test that's as downright dirty (on purpose) as Toia's... rolloff is better in the movie examples out there, but can we see some basic/comparable footage before everyone makes a call?

We're not comparing apples to pears or even oranges (they're not the only fruit)... it's apples to nuts - LOL.

Seriously, I'm perplexed by this backlash... when I see some studio shoots (or controlled lighting) and some R3D's, I will make my judgement - until now, I am excited about this test as I can see what's around the corner.

Maybe it's just me... and the 'Way of the Web' these days.

 

[jake blackstone]

I graded everything in RCX because thats all I had that could actually see the 6k R3D's... Nothing else worked, AE, PS, FCPX, Smoke, CsPREM, resolve... nothing.

Which is what anyone would expect. I'm sure SDK for Dragon is not out to the manufacturers. Once it's out, it will be a week or so to see updates for the Dragon.

So once finished grading in RCX, I exported out DPX sequence and put into Smoke, then AE.. but as we all know, a DPX workflow is slow. I wanted to edit to get this out to you guys quickly... . so i ditched that idea.. then decided to crank out some 4444 PRORES 4k & 2k export tests to see against the DPX's...no difference at all, no gamma shifts like the old days.. all perfect.

DPX workflow is not slower or faster than any other chosen codec. It actually exactly the same. It does, however, requires much faster drives, because, unlike Prores, it's uncompressed.
And using Dragon in no way would affect gamma shift. Frankly, gamma shift between Prores 444 and DPX always was negligible.

I did it in my laptop because I just wanted to see if a 6k work flow was possible on a Laptop and was it going to slow me down when I'm out on location... In short, it doesn't.

Unless you had finished in 6K, which you didn't, your workflow is still good old HD. RCX just scaled it down to HD from 6K. Now, if you managed to finish your project in 6K THAT would be a great trick

 

[Scott Crawley]

I'm a bit confused. Didn't RED promise 20+ stops of DR? But based on Mark's test, it looks like only +1 stop in the highlights and +2 in the shadows? Is the DR expected to change drastically over the next few weeks?

There might have been some pre-production bravado bouncing around but I don't recall reading any promises. This is what I read:

Happy New Year everyone... just a bit of an update on Dragon.


There are some people that think RED has been standing still.. Quite the contrary. We just have been a bit quiet with our hands full harnessing the fire-breathing power of the Dragon.

Dragon was born from a brand new pixel design, with a new fab process and a new read out architecture that has resulted in a sensor that is cleaner and that has more dynamic range than anyone expected.

One of our most reserved sensor engineers wrote Jim and I this morning after shooting a test and these were his words:


" First time in ten years that I'm speechless "


Getting Dragon done was one of the most difficult things we have ever done... but it has turned out to be much more than we could of ever imagined.

This much range coming off of a sensor has never been done before.. let alone at over 6k at 80 frames a second.. so we had to beef up the entire infrastructure of EPIC.

This is likely to affect the price of the upgrade a bit.. and its going to close the door on any chance of a Scarlet to be able to upgrade to Dragon.

We are however.. going to be offering a trade-in program much like we did with the R1 for Scarlet customers to get into an Epic Dragon.


So on the last day of the year of the Dragon, I am going to share with you a frame that was taken this morning.

Don't look at image quality here... this is a dirty debayer with no black calibration, no offset correction and no processing on a special 21 stop DR Chart from DSC.

This is from a prototype " Frankie " Epic camera with a really, really expensive engineering lens that really, really sucks at taking pretty pictures, on a non-sealed lens mount ( hence the light bloom ) but it gives you a bit of an idea on just how powerful all your cameras are about to become.

"Real" Images coming soon... ( but not tonight )

It appears that the Dragon will have about 4 stops more DR and less than half the noise of the current Mysterium-X sensor.

Jim

Mark's estimate of +3 and Jim's estimate of "around 4" are not at odds IMO

 

[Brian Boyer]

Red didn't ask Mark to do a stress test. That was his choice.

All I'm suggesting is that critiquing a compressed Vimeo video just seems odd to me.

As for me, I tip my hat to Red and Crew, then Mark Toia, and carry on.... with my $9,500 patiently waiting to be exchanged.

I know it was Mark's idea. But, it doesn't matter whose idea it was. The first thing out of the gate was a stress test. They're designed to get down to the nitty-gritty. I don't see how people can be faulted for commenting on that level, given the nature of the test.

If all people wanted was to participate in an orgy of praise, a beauty reel might have been a better route for a debut, is alls I'm sayin'. I actually like it when people are critical of RED. I usually learn something, whether they're proved right or wrong and regardless of if I agree or not.

Maybe there should be a new rule...no critical comments for the first 48 hours or the first 48 pages, whichever comes first.

Don't cinematographers the world over rely on Vimeo to showcase their work? Those videos get critiqued all of the time. Besides, Vimeo allows for source file downloads, which is how I watched Mark's video. And aren't Blu-ray discs encoded primarily in H.264 these days, the same as Mark's video?

I know we've been spoiled with raw and 4k+ resolution but how many people will be watching the end result of what Dragon offers in anything better (until UHD takes off, of course)? I don't see Vimeo being quite the limiting factor you suggest.

I, too, tip my hat to RED, Mark and anyone else contributing to the community, but I didn't pledge a blood oath to never raise a concern. And when I (or you) do, it doesn't mean I'm less appreciative than the uber-exuberant among us or betraying RED's sacred trust.

I feel like I thanked RED aplenty by giving them my business when I couldn't really afford it.

 

[Timur Civan]

Out of curiosity, if you were to shoot 5k, would it need special processing as well?

(Read: sneaky way of finding out if its a full on 16bit signal now.... Of course the current RCX wouldn't run it...)

 

[David Battistella]

DPX workflow is not slower or faster than any other chosen codec. It actually exactly the same. It does, however, requires much faster drives, because, unlike Prores, it's uncompressed.

Jake,

uncompressed workflows are actually faster and less processor intesive than ProRES which does require a decompression of the CODEC, uncompressed is a faster, less processor intensive pipeline than working with compressed codecs (H.264 is a great example of a processor intensive codec) Maybe you are talking about data rates, as DPX frames are much larger than compressed frames so you need fast (fiber channel storage or Thunderbolt raids, etc) to push all that data.

David

 

[Phil Holland]

I think what people are missing from this is that when actually lighting, we will rarely create a scene with 15+ stops of latitude.

I've held off posting until now (on this particular section of the debate... but you're right.

Capture is one thing. I have literally thousands of HDRIs where I can measure total available dynamic range on a giving location and lighting condition.

But let me poison minds with this thought. Even if you've lit a scene and it has say 9 stops of latitude. Depending on how you choose to expose for that sliver of histogram you will be able to have freedom to expand that accordian and push that sled around a bit. Or perhaps more practically exposing one way or another to get much more detail in your shadows and darks, or choose to retain your lights and highlights and move them closer to say a midtone. It will be interesting to see where people favor certain exposure choices to maintain a rich texture in low and high key situations. Anywho....

This really isn't about shooting scenes of 20 stops of dynamic range at all times. That's not the best way to look at it. If you are shooting high contrast daylight and exposing into the shadows, not entirely uncommon, you're gonna fill that histogram up with a lot of information.

This is more about having freedom and choice. Freedom to expose how you want and shoot under whatever lighting condition you'd like without breaking your image.

Whenever I get to do some down and dirty testing I'll try to showcase that for sure. The added dynamic range has hefty benefits during capture and in post. That's the whole thing here.

I'm a bit confused. Didn't RED promise 20+ stops of DR? But based on Mark's test, it looks like only +1 stop in the highlights and +2 in the shadows? Is the DR expected to change drastically over the next few weeks?

Initially, when Red did their first dynamic range tests with Dragon they measured over 20 stops straight off the sensor on the chart. It was mentioned by Red as well that the number would be lower once color science was worked out as well as a couple other factors. Hence our new somewhat odd designation of 16+, which is a safe bet at this point. Based on my own calculations, which is mostly theoretical at this point, I'm thinking somewhere between 3.3 and 4 stops more of an improvement of Mysterium-X depending on a couple factors. I'm also betting there's a little more meat in there too, so we could be looking at bit over 4 stops even once it's all said and done. Don't hold me to it though. If it was straight 16 I'd still be pleased as a pig in slop.

 

[Kevin Marshall]

Initially, when Red did their first dynamic range tests with Dragon they measured over 20 stops straight off the sensor on the chart. It was mentioned by Red as well that the number would be lower once color science was worked out as well as a couple other factors. Hence our new somewhat odd designation of 16+, which is a safe bet at this point. Based on my own calculations, which is mostly theoretical at this point, I'm thinking somewhere between 3.3 and 4 stops more of an improvement of Mysterium-X depending on a couple factors. I'm also betting there's a little more meat in there too, so we could be looking at bit over 4 stops even once it's all said and done. Don't hold me to it though. If it was straight 16 I'd still be pleased as a pig in slop.

Even that's not entirely correct. Jarred mentioned that for a wedge to count, it needed to have a wedge on either side. The brightest wedge is clipped, and the darkest wedge is the noise floor. So the most optimistic reading of that chart would be ~18 stops, though as you mentioned Jim and Jarred said it would be more like 16.

 

[Phil Holland]

Even that's not entirely correct. Jarred mentioned that for a wedge to count, it needed to have a wedge on either side. The brightest wedge is clipped, and the darkest wedge is the noise floor. So the most optimistic reading of that chart would be ~18 stops, though as you mentioned Jim and Jarred said it would be more like 16.

Heh. I should have been more specific when I stated "a couple other factors", but yes. Truth.

 

[paul schefz]

This is more about having freedom and choice. Freedom to expose how you want and shoot under whatever lighting condition you'd like without breaking your image.

add 200iso to 4000iso to the 16+ stops DR, add the motion mount to the mix.....i just can't see a situation that can't be handled as long as you kindof know what you are doing...

 

[Robert Ruffo New]

sorry, i may have lead you astray, I meant thats what we were pushing too.. 7 stops ND. Thats when it got to the point were I didn't accept the pictures I was getting.
We were noticing IR starting to come in around 5 stops... not as bad as MX.. but it was still there, Anyways, i don't think the shipped cameras will have the issue. IR will be sorted by then I sure.

Thanks for clarifying! But still, 5 stops = ND 1.8 with no shift. Then again, IR tests would need to be done: A lot of heat/sun onto synthetic fabrics and/or foliage. Not sure how hot it was where you were shooting. On a cold winter's day outside )in Canada) IR has never been too much of an issue, but your results are still encouraging.

 

[Noah Yuan-Vogel]

DPX workflow is not slower or faster than any other chosen codec. It actually exactly the same. It does, however, requires much faster drives, because, unlike Prores, it's uncompressed.

This seems like an odd statement since it sounds like you are saying all workflows are the same speed if you compare them running on hardware with infinite speed, yes that's true but a bit impractical

In the majority of cases, DPX will be a slower workflow than Prores given that DPX workflows are more commonly limited by hard drive speed (its relative limitation) than Prores workflows are by CPU speed (its relative limitation). For realtime decode/encode, just about any modern laptop nowadays has the CPU power to encode/decode a couple streams of prores and any hard drive can handle the bitrates required for several streams of prores, whereas a single 1080p DPX sequence requires 10x more space and at least 2 3.5" hard drives in a RAID0. Of course there is a difference in quality between the two, however. Seems like there would be very few situations in which DPX would be faster to deal with than prores. Probably the only case would be if you have a super old computer hooked up to a really fast RAID? Or is there something I'm missing?