RED Color - A look at REDWideGamutRGB/Log3G10

[John Brennick]

I'm a very inexpensive image release

So Graeme's been cranking away further. Here's some comparisons between DRAGONcolor2/REDgamma4 and the latest REDWideGamutRGB/Log3G10 w/ a Transform LUT for Rec.709.

I'm back in LA now and can test this much more aggressively and have been. This new method is in a pretty good place at the moment. Looking forward to chewing through more Dragon and Helium files in particular, which should line up nicely with an upcoming shoot and camera test.

Overall I'm seeing:
- A thicker/denser look with a smoother highlight roll-off
- More three dimensionality to the images which is much more noticeable on larger screens
- Improved skin tones
- Richer more vibrant hues across the board. Especially in achieving purer blues
- Much better handling of out of gamut/edge of gamut hues, even the tricky stuff as seen in the hellish first image
- Subtle increases in shadow detail
- Slightly less noise in motion tests due to this method

There's going to be a whole lot of conversations in the near future about this process, but it's certainly an improvement from what I can see. It's very flexible in the grade as well.

Jesus these are incredible.. I went back and looked at the originals from the first posts and what a huge leap! Graeme you've outdone yourself. The first, second and last are my absolute favorites. The rolloff is gorgeous and it just makes you realize how much is really hidden inside this raw file to do with what you will. The first one, sooo gorgeous... The feeling this gives me is that this new approach is doing something a bit different than the DC2/RG4 approach.. To me it feels like it's respecting the integrity and fidelity of the colors from the image instead of "forcing it" to separate. I was always more partial to RG3 than RG4 because it didnt crush it as much. However, it was always somewhat of a conundrum. Going RG3 didn't come without its trade off either. RG3 never felt like the colors punched and made me feel in the image. However the subtlety of RG3 was always preferred. Now we come to this new approach.. (by the way, Graeme, did I tell you amazing job? jesus... soo nice..) Looking at the second image this all becomes more mind boggling. The contrast feels a lot more (dare I say) "organic"? It feels to be softer and less aggressive, however the separation and saturation is immediately noticeable. Thats incredibly difficult to do... To smooth out contrast, yet increase perceivable saturation and color separation? Voodoo.... Thanks Phil for putting this together..

 

[Sabovic Adis]

Tastefully dealing with out-of-gamut colours has been the #1 request.

Graeme

Respect, Maestro! :biggrin:

Question:
Is it possible to put the "RWG/LOG3G10/709LUT" combo instead of "709/709" combo in RCXP?

 

[Brody_B]

I'm a very inexpensive image release

So Graeme's been cranking away further. Here's some comparisons between DRAGONcolor2/REDgamma4 and the latest REDWideGamutRGB/Log3G10 w/ a Transform LUT for Rec.709.

Overall I'm seeing:
- A thicker/denser look with a smoother highlight roll-off
- More three dimensionality to the images which is much more noticeable on larger screens
- Improved skin tones
- Richer more vibrant hues across the board. Especially in achieving purer blues
- Much better handling of out of gamut/edge of gamut hues, even the tricky stuff as seen in the hellish first image
- Subtle increases in shadow detail
- Slightly less noise in motion tests due to this method

There's going to be a whole lot of conversations in the near future about this process, but it's certainly an improvement from what I can see. It's very flexible in the grade as well.

This is coming out great! I see sharper images, especially in face paint girl, maybe thats what is being referred to as more 3D. I am also seeing a healthy amount of red, maybe a little too healthy??


Graeme, this is incredible, you are doing gods work my friend. Can't wait to see the official release.

 

[Marc Wielage]

If you just look at the chip chart alone (blowing up Phil's images and applying zero correction), the differences are not gigantic:

DragonColor2-RedGamma4

RedWideGamut-Log3G10



Scopes are not that far off:

DragonColor2-RedGamma4

RedWideGamut-Log3G10


There is a definite difference there, but it's not enormous and it's all stuff I think could be easily tweaked in post. I would be much more interested in seeing how the dynamic range has changed and how the camera handles highlight overload and noise issues.

I would also much rather see tests done with the DSC charts, because the XRite charts are pretty bogus with the specific values they picked for colors in terms of video accuracy, particularly in magenta & green.

 

[Kazu Okuda]

i wonder how redlogfilm 3rd party luts by like Impulz or Koji react to WideGamut/Log3G..

 

[Brian Boyer]

If you just look at the chip chart alone (blowing up Phil's images and applying zero correction), the differences are not gigantic:

DragonColor2-RedGamma4

RedWideGamut-Log3G10

You managed to do something to the images because many of these colors are way off and all of them are more saturated than what appears in the original photo.

 

[Sabovic Adis]

"Color management" tag in .jpg?

 

[Thai Christen]

Colors are richer and pop more. Also light roll off smoother, although subtle. Definitely nicer but something a typical client won't notice IMHO.
Having shot with the Epic-W for nearly a week, what I can say is Helium tends to clip sooner than Dragon. I know, it is still early stage and everything can change once new color sciense is released. For the moment, I am still adjusting on how to expose correctly on Helium since it is very different to Dragon. So, some comarison tests on DR, noise and highlights handling would be helpful.

Thanks Graeme and Phil for the immense work!

 

[Fergus Cahill]

Hey Thai, Interesting to hear what you're saying about clipping sooner. What Base ISO are you using?

Sorry I hope thats not too off topic!

 

[Marc Wielage]

You managed to do something to the images because many of these colors are way off and all of them are more saturated than what appears in the original photo.

Nope, that's just a straight cut and paste and export. Send me an R3D file and I'll be glad to use that.

Again, the biggest issues here are: 1) it's a bogus chart, 2) the chart isn't full-frame, and 3) we don't have a full-res image to work with. Shooting a chart that's waaaaaay off to the side is not nearly as helpful as making the image fill the frame.

It also points to the issue of still frames posted to a web page. You can't make any real judgements, particularly in an environment like this with (mostly) uncalibrated monitors. I don't doubt there are bigger differences that could be seen in real tests with uncompressed files, real monitors, scopes, and moving images.

 

[Hrvoje Simic]

I would also much rather see tests done with the DSC charts, because the XRite charts are pretty bogus with the specific values they picked for colors in terms of video accuracy, particularly in magenta & green.

Regardless of which chart you use, color chart analysis in examples like these works for a particular scene color adaptation for color grading purposes and just a general impression.

For judging camera performance that approach doesn't work.

 

[Thai Christen]

Hey Thai, Interesting to hear what you're saying about clipping sooner. What Base ISO are you using?

Sorry I hope thats not too off topic!

Hey Fergus, here is a thread regarding this topic
http://www.reduser.net/forum/showthread.php?148722-DRAGON-or-HELIUM

I use 1280 with Skintone/Highlight, 1600 with Standard OLPF. Haven't tested lowlight OLPF yet, but guess 2000.

For the moment, I expose below IRE 92-94 on Helium (with LUT or DragonColor and Redgamma3) and don't worry about the shadows because noise is a tag cleaner on Helium than Weapon. As soon as you push beyond IRE 94/95, even though there are no light stop popping up in the histogram, when open up in RedCine, highlights look like they are "clipped". Of course you can retain highlights (they are still there), but I rather have it right in camera than tweaking in post.

 

[Fergus Cahill]

Hey Fergus, here is a thread regarding this topic
http://www.reduser.net/forum/showthread.php?148722-DRAGON-or-HELIUM

I use 1280 with Skintone/Highlight, 1600 with Standard OLPF. Haven't tested lowlight OLPF yet, but guess 2000.

For the moment, I expose below IRE 92-94 on Helium (with LUT or DragonColor and Redgamma3) and don't worry about the shadows because noise is a tag cleaner on Helium than Weapon. As soon as you push beyond IRE 94/95, even though there are no light stop popping up in the histogram, when open up in RedCine, highlights look like they are "clipped". Of course you can retain highlights (they are still there), but I rather have it right in camera than tweaking in post.

Thanks Thai! Really interesting, I will have to catch up on that thread. 94/95 IRE is really low. Will be interested to see how that develops.

 

[Thai Christen]

Thanks Thai! Really interesting, I will have to catch up on that thread. 94/95 IRE is really low. Will be interested to see how that develops.

I know. This is why I am curious about other Epic-W shooters on how they are adjusting the situation.
And of course what magic the new color science can do.

 

[Stacey Spears]

Again, the biggest issues here are: 1) it's a bogus chart, 2) the chart isn't full-frame, a

Marc, when shooting a test chart, how much of the frame would you like it to fill? e.g. I would like to include a test chart at the start of each interview. Does it really need to fill the full frame? Can it fill 50% of the frame? Lets say I want the interviewee to hold the chart in front of them, just below their face.

 

[Joel Arvidsson]

Is this the same version of REDWideGamutRGB/Log3G10 as in the latest release version of redcine x pro?
Is it correct that REDWideGamutRGB/Log3G10 is implemented in fcpx and davinci but the original version?
Wish to se a adobe update very soon with this goodies in!

 

[Brian Boyer]

Nope, that's just a straight cut and paste and export. Send me an R3D file and I'll be glad to use that.

Again, the biggest issues here are: 1) it's a bogus chart, 2) the chart isn't full-frame, and 3) we don't have a full-res image to work with. Shooting a chart that's waaaaaay off to the side is not nearly as helpful as making the image fill the frame.

It also points to the issue of still frames posted to a web page. You can't make any real judgements, particularly in an environment like this with (mostly) uncalibrated monitors. I don't doubt there are bigger differences that could be seen in real tests with uncompressed files, real monitors, scopes, and moving images.

Um, I don't know what to tell you. The colors in the example you posted are off...big time.


If the colors in your image...

...look the same to you as the colors in this photo,

one of us is doing something terribly wrong.

Looking at the magenta swatch (F2), the purple one (G4), the lavender (H2), red (F4) or the kinda dusty rose (G3), on my monitor (calibrated), viewed in three different browsers and Photoshop, the colors aren't right.

Perhaps the difference is in what color space you saved/handled the images. Or, maybe the service hosting the photos somehow altered the color. Whatever the cause, I don't know why you're not the seeing the shifts as plainly as I am.

Is anyone else seeing Marc's images with the correct colors?

 

[Mike Nagel]

Marc, when shooting a test chart, how much of the frame would you like it to fill? e.g. I would like to include a test chart at the start of each interview. Does it really need to fill the full frame? Can it fill 50% of the frame? Lets say I want the interviewee to hold the chart in front of them, just below their face.

hold it right in front of their noses - so the chart should still be in focus and lit the same way as the subject... use the same lens, so that will be a much smaller chart on wide shots, but that's just how it is...

make sure the black is not reflecting on the DSC chart... keep part of the face in, so u can actually see the skin tone of the face when u neutralize color in post... yes, u will get some not-so-welcome surprises on some RED footage depending on ur lights, even with the charts...

the point of getting a larger shot of the chart is to get a larger sample area, when u align vectors in post...

 

[Hrvoje Simic]

u will get some not-so-welcome surprises on some RED footage depending on ur lights, even with the charts...

As with every other camera, when specific light properties (especially mixed light) meet specific camera response properties.

 

[Stacey Spears]

hold it right in front of their noses - so the chart should still be in focus and lit the same way as the subject... use the same lens, so that will be a much smaller chart on wide shots, but that's just how it is...

Thank you Mike, that is what I had in mind.

make sure the black is not reflecting on the DSC chart...

On my chart, I have the version with the hole in it and the velvet lined cave that mounts on the back of the chart. I think they call it CaviBlack.

Is this the same version of REDWideGamutRGB/Log3G10 as in the latest release version of redcine x pro?

The latest SDK has an updated Log3G10. This was released last week. This matches the Log3G10 in the official release of build 42, which also came out last week. There was even a new camera FW update.