Dragon compared to 35mm Vision3 200T

[Matt Ryan]

https://vimeo.com/88834695

Most likely shot with the original OLPF, so chances are it would fair even better on highlights. I'd be curious to see the Dragon footage re-processed with Dragoncolor as well.

P.S. this is not a test I was involved in, just sharing.

 

[Nick Morrison]

I loved this test....

 

[Patrick Tresch]

Thanks Matt for the test.

I liked the filter test. I would try to use the Dragon on a higher iso level to get some noise and fine tune the noise color in post to keep some subtile BW noise texture in the R3D picture. When doing this the exposure has to be right on.

I apreciate you sharing this work. Cheers to all contributors.

Patrick

 

[Matt Ryan]

Hey Patrick,

I wasn't involved in the testing, just sharing. Anyways, I'd probably just use filmconvert to match the grain vs. bringing out noise when shooting. But I prefer to have a clean start myself.

 

[Phil Holland]

Posted this in another thread this morning, but it's good that it has it's own thread.

This pretty much mimics what I've been saying and showing since last year. Red has really tackled hard the apparent digital-ness of the image coming out of Dragon.

The image is far more organic looking. Clean. Meaty. Colorful.

Pretty much going back to the initial promise of a "digital film alternative". Now with Dragon's Dynamic Range, Highlight Roll-Off, Color Science, Format Relevance, and various other features I truly do feel we have the first true "digital film replacement".

It's really, really good.

 

[Ivan McCullough]

Very interesting and well done test.
Dragon looks a bit green or yellow/green in comparison to the kodak, likely an easy thing to time out.

Overall though very impressive highlights and tonality.

 

[Matt Ryan]

Very interesting and well done test.
Dragon looks a bit green or yellow/green in comparison to the kodak, likely an easy thing to time out.

Overall though very impressive highlights and tonality.

That's a combination of the old olpf and the old Redcolor 3 color science. New OLPF gets rid of the green cast.

 

[Hans von Sonntag]

Those tests are always flawed because they use different gamma curves or/and low saturation. The skin tone of the lady has a porcelain character. This way you cannot tell how rich colour diversity in normal skin tones are. Because all is bright and even exposed you cannot judge how rich the shadows in the skin tone are.

Noise, of course, can be tested well. No wonder that Dragon outperforms film here. I would doubt that counts also for colour richness. Nonetheless Dragon is a milestone. Tomorrow mine will arrive. Haha!

Hans

 

[Robert Ruffo New]

Hans, pale skin is the hardest. On Mx she would be close to white all else being equal.

 

[Gavin Greenwalt]

Yeah I am eager to see a debayer without the green cast someday. It's been dogging RED for half a decade now. Was really hoping we could put that behind us completely with Dragon. I highly doubt it's the OLPF though. I can't see a reason why colorscience couldn't correct it though. So hopefully DragonColor/DragonColor?2? nails it once and for all. It also looks like it would be tricky to time out since much of her skintone is sufficiently warm, but other parts aren't so it would take some form of secondary or 3DLUT.

 

[Burke Doeren]

Gavin, there was recently a thread about raising the saturation pre-debayer vs post-debayer, and the results showed that the greenish cast is drastically reduced when you raise the saturation post-debayer. You can do this in RCX in the lower effects panel (near alchemy, film grain, etc) and see the results for yourself. Plus, DragonColor seriously helps smooth everything out, especially at high saturation levels.

 

[Vithya Sharma]

Very interesting ....

 

[Marc Wielage]

That is a good test, and whoever color-corrected it matched the footage very well. I would rather see a test with 5219 (500 ISO), which is much more-widely used for film projects and is closer to the Dragon's exposure range. I don't dispute that 5213 (Vision 200) is finer grained, but I honestly like the blacks better in 5219.

 

[Gavin Greenwalt]

You can do this in RCX in the lower effects panel (near alchemy, film grain, etc) and see the results for yourself. Plus, DragonColor seriously helps smooth everything out, especially at high saturation levels.

I know, read the signatures in that thread carefully. Post gamma saturation certainly helps lessen the effect but regardless of where you apply saturation in the color pipeline the green skew is a hue characteristic not a saturation artifact. Increasing saturation in linear space simply amplifies an existing characteristic. If you crank up the saturation in the film plate it will remain warm--just more intense. The grade is closely matched enough in this test that I don't expect different RAW settings to radically change the image. It's an excellent comparison in my opinion of the more subjective attributes of film and RED. I think it also very clearly illustrates why people still rave about film (in spite of the gateweave).

 

[Brian Boyer]

I know, read the signatures in that thread carefully.

And I thought I was going to be the one dogged by that thread. I guess no good deed goes unpunished.

Yeah I am eager to see a debayer without the green cast someday. It's been dogging RED for half a decade now. Was really hoping we could put that behind us completely with Dragon. I highly doubt it's the OLPF though. I can't see a reason why colorscience couldn't correct it though. So hopefully DragonColor/DragonColor?2? nails it once and for all. It also looks like it would be tricky to time out since much of her skintone is sufficiently warm, but other parts aren't so it would take some form of secondary or 3DLUT.

From what I understand, the new OLPF was accompanied by a color science tweak to match it and the new sensor calibrations. People often say "new OLPF" as shorthand for the whole shebang of hardware and software changes that get done when the new OLPF is swapped in, or, put in, at this point.

That suggests you shouldn't be able to just apply DRAGONcolor to this footage and get the same result you would if it were shot with currently shipping cameras, as I understand it. So, it could be the case that the green cast has been significantly lessened, if not eliminated entirely.

There aren't many .R3Ds out there yet of skin tones from the new OLPF that haven't been shot under crazy lighting conditions but the few I've seen of normal lighting don't seem to have a green cast problem, even without DRAGONcolor.

Then again, they aren't side-by-side with a film frame of the same shot for comparison.

 

[Tony Harris]

Nice video but I think trying to compare video to film is a waste of time. Film will never look like video and video will never look like film.
35mm film is Outstanding! RED has made video Outstanding! RED has put an end to cheap looking 16mm film and gave us a greater choice that's close to 35mm film.
Call it cine 35mm video or whatever but I think they both stand on their own. Both are expensive as hell but we have a voice and a choice now. I like the RED vision and what they are doing, hell, call it RED VISION 3 for the Dragon and RED VISION 4 for the next camera. Film making has become sweet on both ends.

Its kind of silly RED trying to copy film and other companies trying to copy RED. Red has given us the power over what we create and film takes that away in the processing unless you have the money. To shoot what I want in film would take two film cameras, normal and high speed. If I shoot with RED, one Epic X will do the job for me and look outstanding.

 

[Ivan McCullough]

Those tests are always flawed because they use different gamma curves or/and low saturation. The skin tone of the lady has a porcelain character. This way you cannot tell how rich colour diversity in normal skin tones are. Because all is bright and even exposed you cannot judge how rich the shadows in the skin tone are. ....

Hans

All tests you do not do for yourself and for exactly what you want to see are flawed.

Its nice to have a peek at what someone else did for themselves, within the boundaries of what that person wanted to test, and make what judgments you can from that. Thats all.

 

[Gavin Greenwalt]

Nice video but I think trying to compare video to film is a waste of time. Film will never look like video and video will never look like film.

Never say never. I would say the two are already very close. FilmConvert often looks very much like film. I don't see how it's a waste of time either. Film is just a photochemical LUT between reality and a subjective perspective on that reality. So are digital sensors. Kodak spent 100 years working on their LUT and I think it looks fantastic. Why wouldn't I want my camera to look like Film just because it uses a different approach to recording photons? To me that would be like arguing "Why would we want an electric car to look like a Ferrari, ICE is ICE and electric is electric?" Because they look sexy as hell would be my response!

 

[Marc Wielage]

I think it also very clearly illustrates why people still rave about film (in spite of the gateweave).

No gate weave with pin-reg scans.

Kodak spent 100 years working on their LUT and I think it looks fantastic.

Small correction: Kodak didn't come up with the concept of a film LUT until 1992, when Cineon was introduced. Mitch Bogdonovich and Glenn Kennel were among the team that made that a reality, and the LUTs were an ongoing development even when I was there in 2001-2004. And even Kodak's own LUTs were not 100% perfect, nor could they necessarily take electronic images and make them look like film.

What they could do is predict how the images would look when printed on film, but even there, there's a lot of potential issues that change the look. I personally think apart from grain, 90% of what's in these "look LUTs" can be done by a good colorist who knows what he or she is doing. I also think the real look of film is ephemeral and based on a hundred different factors, much of which boils down to great lighting and exposure. There is no magic LUT that will take crap digital footage and make it look like great film. And I think a lot of the LUT's reaction will depend heavily on the nature of the material. Not everything translates well.

 

[Gavin Greenwalt]

I'm using "LUT" in the ephemeral/spiritual sense. Photons come in -> Chemicals Reaction -> Pictures come out. It's not a literal table of values but Kodak has spent the better part of a century creating their interpretation of reality or "look". Each film stock interprets reality in a very specific way, as does each digital sensor. I prefer most film stocks' interpretation. DragonColor is a "LUT". Vision3 is a "LUT". Kodachrome is a "LUT". They're all consistent color transforms. Feed them an input color of light or RAW data and they give you a consistent(ish) response. Lens coatings are a LUT, filters are a LUT. All can be modeled as an I/O pairing. Feed a CTO filter 5600ish white light and it'll output 3200ish k light. As far as I'm concerned that's just an optical LUT. It's functionally equivalent to a .cube file that does the same thing just operating on photons instead of bytes.

I think the argument that colorists could match film is similar to the argument that animators can create a photorealistic human animation. It's certainly theoretically possible but it's done so infrequently that clearly the process and technology has work to be done.