Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

New RED Color Science - Test Pilot Feedback Thread

just out of curiosity it almost seems like you shifted the native color balance of the camera by 400 Kelvin +/- a few...

Does this mean that the camera is no longer 5K Kelvin anymore?

In fact i am very impressed with the speed at which this build was released, unseen
since we had been in a beta of 18 for months, which was centered around anamorphic.

could this have anything to do with all the new Digital Cinema comparisons that are being done at the moment? especially the ASC shoot out?

It seems like all the recent tests would be now invalid because you remapped a native color temp for the sensor and reduced the blue channel noise?

personally and this is just my opinion... subjective.... it looks a little warmer.
But then again i had been shooting all my day exteriors at 7000 Kelvin
because it seemed inaccurate at 5600.

is this not accomplishing the same thing but shifting everything a few hundred degrees.
Just like Nate had demonstrated earlier a few posts ago?

Im all down with funky AI's but as a DP id like to know what my native kelvin is now? did it change???
and how this will change if any my current lighting temps.

will this now handle mixed color temps better, because up untill now the camera didnt at all,
and it wasnt a bad thing it let me get some great effects before... but i new that ahead of time and
compensated for it, or played it up as needed.

will that now change ?? if it does, then how will it be different. ?

This is one of my tools, like many others here...

Id love to know how this effects me... and what is now different.
and no offense but "funky AI" is not something i can tell my gaffer to
compensate for when trying to achieve a certain look..

most companies release a white paper when a major change is done... will RED or yourself?

I admire your drive to keep improving your product, i love all that has been done its amazing...
but im confused............

But most of all i wish i didnt have to follow a long thread to try and find my answers.
 
Nate, unfortunately, you can't compare the old and new with the same kelvin color temp, because I increased the accuracy of the calibration of the color temperatures, so that now the color temp that the camera awb sees for the light in the scene or the picker in RA! is very close to the number you'd get by using a color temp meter. Before, the numbers were a touch rough...

Graeme
Click to expand...

If I understand correctly, the native color temp of system has not changed, the Kelvin reference point for a given white balance has changed in the software and now matches the Kelvin readout of a color temp meter more accurately compared to human vision.

I would say treat it like a new film stock and shoot some tests to determine your new reference points.
 
Ernesto, as I understand it the sensor's native Kelvin temperature for balanced colour channels is still 5000K. This is a physical property of the sensor's silicon and has nothing to do with the post-processing of the image.

Now though the new colour science means that the auto white balance, and along with that the detail retained in the blue channel especially, means that you see more tonally accurate colour on the monitors and in the post apps when they auto white balance and debayer the RAW image. The other effect of the new debayering is greatly reduced blue channel noise, which is the second time Graeme has managed to significantly improve this most difficult of all the channels....

HTH

Paul
 
Ernesto, it's just that it's taken me over a year to write the AI code, and that's why we're implementing it now. It's not trivial stuff, and as well as writing the code for the post production software, we had to put it in camera too, so that the metadata is properly meaningful across the two.

Graeme
 
OK,

Got my camera back from RED so now I can play with the new builds too. For the testing, should we still continue to use the test pilot build 18 on the camera? Or should we load the new public beta build 20 for "official" tests?
 
Graeme,
i think its incredable that your code keep evolving
and we keep benefiting. And I a appreciate it..

I also understand the physical limitations of the sensor are at 5,000
but as far as the image i was getting out of the camera
it never looked so.

Did a dirty test with skin tones appear more accurate,
and the on the fly debayering if much improved.

Apart from that I don't understand the rest,
and I don't beleave many people here do either.
At least I'll admit ignorance , but request actual information.

Ofcoarse we will shoot tests and so will everyone else I know.
Just like when kodak would hand out 400' rolls to us for testing

but even then it would come with some sort of liturature to
tell you what to look out for, such as a. Extra stop in your highs
or more contrast or reduced grain because of T-grain.

It's just frustrating is all I'm saying. That we are on page 29 on some thread
and no one here really knows what's going on.
Because it's not been explained, I'm geting tired of people making
assumptions into facts here. It's not accurate or scientific and anything
other then that is just hype.

I love hype too... But I also love facts.
 
Graeme has worked for months on this and the results speak for themselves. We put his new color science out to a group of test pilots so they could use, analyze and report what is different/better and tips for use. It isn't in our best interests that Graeme say exactly what he did to achieve better results.

Jim
 
Jannard said:
Graeme has worked for months on this and the results speak for themselves. We put his new color science out to a group of test pilots so they could use, analyze and report what is different/better and tips for use. It isn't in our best interests that Graeme say exactly what he did to achieve better results.

Jim
Click to expand...

I know all of this will help future cameras, but will the other cameras still have to be worked out individually?
 
Joseph Ward said:
I know all of this will help future cameras, but will the other cameras still have to be worked out individually?
Click to expand...

All the knowledge transfers forward.

Jim
 
Huh Graeme reading all your back posts, I have some idea of what you're doing, and if thats the case jesus. That kind of color adaptive system could have rather large implications and not just for Red. Hope you have patents.
 
I'm sorry for the misunderstanding,
I do not care so much for the how it works
as I am the "what it does, and why" or a simple
"what changed"

is the to much for a humble customer to ask?

I'm a customer not a competitor don't be on the defence.
Were all on the same team.
with a product he owns and promotes?
 
I concur with Ernesto about the allowing technical folks like D.I.Ts who have to explain to the countless DPs they work with what the new" Color Science" is. The Key term here is "science".

I sign NDA's all the time with millions of dollars in damages. If that's what it takes to know the science sign me up!

Congrats to Red and Greame for your hard work and true artistry.
 
My impression of what is happening is that you're using the blue samples less for luma information now in order to mine the cleaner red and green channels for all they're worth before resorting to the blue raw data? A sort of weighted debayer. But then with a bunch of intricate algorithms to ensure this doesn't result in a loss of blue chroma or incorrect luma.

As an apparent result this would shift the color temperature to be slightly warmer.

In other words this is like shifting the color temperature on the sample side but still maintaining correct colorimetry through some dark magic.

Is this a good way to think about it?
 
Gavin Greenwalt said:
My impression of what is happening is that you're using the blue samples less for luma information now in order to mine the cleaner red and green channels for all they're worth before resorting to the blue raw data? A sort of weighted debayer. But then with a bunch of intricate algorithms to ensure this doesn't result in a loss of blue chroma or incorrect luma.

As an apparent result this would shift the color temperature to be slightly warmer.

In other words this is like shifting the color temperature on the sample side but still maintaining correct colorimetry through some dark magic.

Is this a good way to think about it?
Click to expand...

"Dark magic" is the best explanation I have heard so far... :-)

Jim
 
So far, my own assessment is that the camera by default is now warmer (for a given K setting). Also I see blue channel is being rendered better, but the warmer channel balance makes it difficult to figure out just how much, as that makes the blue channel/overall noise level better as well. I understand that what's visible in the blue channel might be improved in the other channels as well.

In addition, the noise in a debayered image seems to be more monochromatic now, which is very nice.
 
I'd like to pose this question again, as I don't feel like it really got answered; Based on what I'm seeing, we're seeing less noise because the blue channel is turned down in strength to make for accurate white balance temperatures. (If this is not the case, please correct me).

If the blue channel is turned up to the same strength to match what an earlier build would look like, the noise is about the same then?

This question is based off the images Nate posted a few pages back...
 
I'm not the authority on this, but from my experience with this build, there's a lot more going on than just turning down the blue channel. The blue channel itself is much cleaner (whether or not it's turned up to match the levels of the previous build).

I can shoot something at ISO 1000 and match the levels and color balance using this build and the previous build, and the new build has dramatically less noise. ISO 1000 wasn't usable without noise reduction before this new build and now it is (in certain circumstances).
 
You must log in or register to reply here.
Back
Top