Strange-clip-Monstro

[luigivaltulini]

sorry maybe I'm stoning myself ..
Am I doing anything wrong?
Lately I've noticed that recording in RAw at a temperature of 3200 degrees and then increasing up to 5600 I'm losing dynamics.
I can understand if I move 5,000 degrees, but here it happens already after 500 degrees up.
I noticed this on people's faces, so I took a spot test against the wall.
It does not seem normal, normally if I change temperature it should not change the dynamic so drastically.

https://drive.google.com/file/d/1eAnO8f1wdHW4Z9_VoWCwHF4poLRdf-tK/view?usp=sharing

[/url]red by luigi valtulini, su Flickr[/IMG]

 

[Aaron Lochert]

The behavior of the highlight extension algorithm is a little wonky at different color temperatures..

We learned this from this thread:
http://www.reduser.net/forum/showthread.php?157839-Helium-IPP2-Highlight-Test-Huge-improvement

 

[jake blackstone]

Don't use camera metadata color temperature adjustments. Instead, use your primary offsets to reduce exposure and then use color temperature adjustments also using Resolve controls. No clipping...

 

[rand thompson]

Luigi,

Also, part of your problem is that you are using "Rec709" for Color Space instead of "RWGRGB" and BT1886 for "Gamma Curve" instead of " Log3G10" under the IPP2 " Color Science". You transformed your R3d file in IPP2 into Rec709 and Bt1886 at
the import of the R3d file into your timeline instead in the last node of your grade. So you are doing about the same thing as trying to push an already Pre-graded ProRes file.

If you pushed the temp inside IPP2 with the IPP2 Output Transform Lut in the last node of the grade you would get better highlight retention/recovery from the highlight recovery Algorithm.

 

[Aaron Lochert]

Luigi,

Also, part of your problem is that you are using "Rec709" for Color Space instead of "RWGRGB" and BT1886 for "Gamma Curve" instead of " Log3G10" under the IPP2 " Color Science". You transformed your R3d file in IPP2 into Rec709 and Bt1886 at
the import of the R3d file into your timeline instead in the last node of your grade. So you are doing about the same thing as trying to push an already Pre-graded ProRes file.

If you pushed the temp inside IPP2 with the IPP2 Output Transform Lut in the last node of the grade you would get better highlight retention/recovery from the highlight recovery Algorithm.

That is false.

The temperature control in the raw tab is still happening prior to the output transform in the raw tab. It will make no difference. In fact, if you do set your raw tab to RWG and Log3g10, you can watch your waveform with NO output transform and still see that you can still truncate the tops of peaks off depending on the color temperature set in the raw tab. It's happening in the SDK's white balancing section, which is why Jake is suggesting to avoid the color temperature slider and instead use Resolve's controls to change color temp.

RWG/Log3g10:

 

[David J. Buchanan]

What the hell kind of weird bug is that? Is this a RED issue or a Resolve issue?

 

[David J. Buchanan]

Also, shouldn't this be in the Monstro thread?

 

[rand thompson]

Aaron,



You Are correct. It makes no difference in Resolve. But a big difference in Redcine-X ,Which is exactly why I only change the color temp in REDcine-X. I setup an IPP2 RWGRGB/LOG3G10 project in resolve to take in those settings from Redcine-X and correct everything else in Resolve. The colors on the chart are fucked but the primary goal of this test was to recover the blown out right side of your face. This was to show how temp and tint in redcine-x makes a difference and not so much in resolve for Raw.

 

[jake blackstone]

What the hell kind of weird bug is that? Is this a RED issue or a Resolve issue?

It is not a Resolve issue, as all color grading software platforms use the same Red SDK. And that is why I just verified, that Baselight also exhibits an exactly the same behavior when using metadata adjustments.
Personally, I haven't noticed this behavior earlier, as I almost never use metadata adjustments. It is much easier to do all necessary adjustments using software controls, with only possible exception being the exposure.

 

[Aaron Lochert]

What the hell kind of weird bug is that? Is this a RED issue or a Resolve issue?

It's a RED phenomenon; Specifically the highlight recovery algorithm. It is more or less effective at various color temperatures. It happens in RCX as well as Resolve. See below.

You Are correct. It makes no difference in Resolve. But a big difference in Redcine-X ,Which is exactly why I only change the color temp in REDcine-X. I setup an IPP2 RWGRGB/LOG3G10 project in resolve to take in those settings from Redcine-X and correct everything else in Resolve. The colors on the chart are fucked but the primary goal of this test was to recover the blown out right side of your face. This was to show how temp and tint in redcine-x makes a difference and not so much in resolve for Raw.

That's not at all what I'm saying.

What I'm saying is the inconsistent recovery of clipped peaks at various color temperatures issue is a RED SDK phenomenon that happens prior to any output transforms. It happens in RCX as much as it happens in Resolve. Set your IPP2 monitoring in RCX to view straight RWG/Log3g10 and mess with the temperature slider and you'll see peaks come and go as you slide that slider around. Your worklfow of only changing the color temp in RCX isn't gaining you anything over doing it in Resolve -- You're creating unnecessary extra work for yourself.

Here's Resolve and RCX side by sides at two color temperatures with both RWG/Log3g10 and Rec709/BT1886 (medium contrast/medium roll-off) versions. You can see that they look the same across both software when the settings are the same, output transform or not. Resolve version always on the left of the RCX version:

So telling Luigi his issue is because of when in the workflow he's applying the output transform is irrelevant to the discussion.

It is not a Resolve issue, as all color grading software platforms use the same Red SDK. And that is why I just verified, that Baselight also exhibits an exactly the same behavior when using metadata adjustments.
Personally, I haven't noticed this behavior earlier, as I almost never use metadata adjustments. It is much easier to do all necessary adjustments using software controls, with only possible exception being the exposure.

The interesting thing is that even if you avoid adjusting the metadata, you might not avoid the issue because it's counter-intuitive. The example I've been using of my mug was shot with a 5600K balanced fixture. The metadata color temperature that yields the best highlight recovery in this scenario is somewhere around 3200K. But of course I shot it at 5600K and would of course deliver it to a colorist with that metadata. But a colorist would have better luck actually changing the color temp counter-intuitively to 3200k and then timing that color back to 5600K in their grading software.

 

[rand thompson]

Aaron,


I understand what you are saying. I noticed that about 6 or 8 months ago when first trying to recover some overexposed r3d files. it seems the cooler you make the image, the more recoverable the highlights are. But I believe I remember RED mentioning something awhile back about the highlight recovery algorithm having something to do with the "Blue" Channel of the RGB image.

So it would make since that the algorithm would exhibit better recovery characteristics at different color temperatures. The problem is what it is until Red Explains why this is happening or if its even a problem at all.

 

[luigivaltulini]

Don't use camera metadata color temperature adjustments. Instead, use your primary offsets to reduce exposure and then use color temperature adjustments also using Resolve controls. No clipping...

Thanks Jack,
normally my colorist doesn't use metadata, yep this is best way.

This thing came out during a short film where at a certain point we wanted everything to be warmer .. and at that point I saw the faces deteriorate on the faces and some parts of the frame .. I thought it was a problem of some Davinci setting.
However I don't like this thing ...

There is no way to correct, you have to stay 1 or 2 stops less light, or be sure to stay in a temperature range no more than 500 degrees above or below, especially in the most illuminated parts.

Thanks Aaron e Rand

 

[rand thompson]

You're welcome Luigi! Maybe the Red Team or Phil can shed more light on this.

 

[rand thompson]

Luigi,

Something else I learned maybe longer than 8 months ago and I believe it was from red files you posted last year. Whereas Monstro only needs a certain Temp. To be more recoverable, Helium and Gemini need both Temp and Tint to be equally as recoverable as Monstro. Again, maybe Graeme or Phil or another member of the Red Team can explain why this is.

 

[Jake Wilganowski]

So basically you have more dynamic range at a very particular white balance and tint, this is good thing?

Or stated another way, Red loses dynamic range as you stray from a particular white balance and tint? Can this "added" dynamic range ever be seen WITHOUT this post process?

 

[rand thompson]

Jake,

Yes but you have to start from a warmer temp like around like 5000 to 5600k and go towards around 3000 to 3200k.However, If you shoot at 3000 to 3200k and over expose an image there will be no way to recover it by going warmer towards 5000 to 5600 , it will only make it worse. This is what I think Luigi has discovered.

 

[jake blackstone]

Jake,

Yes but you have to start from a warmer temp like around like 5000 to 5600k and go towards around 3000 to 3200k.However, If you shoot at 3000 to 3200k and over expose an image there will be no way to recover it by going warmer towards 5000 to 5600 , it will only make it worse. This is what I think Luigi has discovered.

No, as I said, if you don't use metadata adjustments, there is no limitation in primary grades regardless of color temperature direction. I speculate, Red SDK works in this order- first Color Temp and then Exposure. In that case, some of the color channels would get clipped before you can adjust the exposure. But by using Primaries, you can first reduce the Exposure and then adjust Color Temperature without the clipping. Regardless, there is ZERO reason to use metadata adjustments, as it was clearly shown by FilmLight. In the video they use Alexa RAW material, but the same principal still applies to Red RAW. Baselight can easily duplicate all metadata adjustments with the Basegrade operator. Resolve too can perform similar thing, just not as elegantly, quickly nor precisely...
https://vimeo.com/251798431

 

[rand thompson]

Jake,

I was actually referring to Jake Wilganowski`s post. But yeah, you would know much better than I as you are the seasoned colorist in this discussion. So if I am understanding you correctly, when you say Color Temp you are referring to Primary color correction and not Camera Raw "Temp". Also , and correct me if I'm wrong, you're saying it's best not to use Camera Raw`s Temp and Tint to correct for clipping in a R3d file ? I saw the video you posted which indeed uses Alexa footage , but from trying to recover both from overexposure, The Alexa files were easier to recover in the method you suggest than the red files

If Aaron or if he would allow me to post the +5 stop overexposed Red Raw file, would you have the time to show the result of your methodology to remedy this problem? I tried every traditional conventional approach "I" could think of to solve this problem. What I used to recover Aaron's clip on the first page was the best "I" could come up with.

 

[jake blackstone]

Jake,

I was actually referring to Jake Wilganowski`s post. But yeah, you would know much better than I as you are the seasoned colorist in this discussion. So if I am understanding you correctly, when you say Color Temp you are referring to Primary color correction and not Camera Raw "Temp". Also , and correct me if I'm wrong, you're saying it's best not to use Camera Raw`s Temp and Tint to correct for clipping in a R3d file ? I saw the video you posted which indeed uses Alexa footage , but from trying to recover both from overexposure, The Alexa files were easier to recover in the method you suggest than the red files

If Aaron or if he would allow me to post the +5 stop overexposed Red Raw file, would you have the time to show the result of your methodology to remedy this problem? I tried every traditional conventional approach "I" could think of to solve this problem. What I used to recover Aaron's clip on the first page was the best "I" could come up with.

All I am saying, for MYSELF only, I find the manipulation of the camera metadata usually is unnecessary.
As far as the overexposure goes, again, for MYSELF, I find once Red material is grossly overexposed, it is pretty much impossible to recover. You may use a bit of Baselight's Highlight Texture too make clipped highlits to look less harsh, but that will not help in bringing those highlights back. On the other hand, Alexa almost never clips the highlights, even when shot on Prores. There is no special methodology in dealing with clipped highlights, regardless of grading platform, once it's clipped, it stays clipped...

 

[Adrian Jebef]

When you change Color Temp / Tint you basically change RGB channel gain settings from camera native which in turn changes your output image values ie: exposure. To state it another way, WB/T settings are directly related to exposure under colored light sources. Even though WB/T setting can be changed in raw the RGB channels are locked in exposure values so all you are doing is adjusting the ratios of the channels. If you have no more over/under range on your channel limits you will get over/under exposure when you adjust WB/T. This is what Luigi ran into. Although all digital cameras exhibit this behavior Red Cams seem to be far more susceptible to it in certain ways.