Blackmagic Forum

There is a problem with Overlay composite mode in ACES in DR16 and 16.1.

Adding film grain by a ProRes scan using Overlay in ACES has never been an issue before. Now in DR16 using overlay results in a major gamma shift. Please refer to stills attached.

What do you mean by overlay? Is that a video clip with alpha?

I think overlay composite mode means it is a composite mode named overlay. Overlay does not need alpha, just as plus, screen etc don't.

As time goes, I find less and less ACES related stuff that actually works.

Jim Simon wrote:What do you mean by overlay? Is that a video clip with alpha?

It's a common way of adding film grain to a shot by having a grain scan on a second video layer then blending it in by the use of Overlay Composite Mode.

Got it. Thanks, Boris.

I am seeing the image darken when using the Overlay mode in Composite in Studio 16.1b1 for Windows. Though I haven't sufficient experience to know what's normal here.

So disappointed to see that this issue has not been fixed with the release of DR16.1b2.

It is making things even more complicated when working in ACES. I am unable to see the grain in context while coloring and have to export to REC709 in order to apply grain.

This is far from ideal, especially knowing that all was good in DR15.

Would someone from BMD please acknowledge the issue.

Thank you!

Strictly speaking, Overlay is an invalid blend mode in any scene-referred color workflow. It doesn't respond properly to values above 1.0.

That said, I can't think of any reason it should have changed from one version to the next, and some clarification of the operation that's being performed would be appreciated.

In the meantime, a 'more correct' application of grain is to subtract 0.5 from the grain plate, then add it to the image. This must be done in floating-point, which should be a given if you're in ACES.

On Resolve side working space is most usually cc or cct log where values don't exceed 1.0, so none of the (scene)linear or pure gamma logic and blend modes properly apply anyway. But undocumented changes from one version to another is a "feature" one seems to have to live with.

Hendrik Proosa wrote:On Resolve side working space is most usually cc or cct log where values don't exceed 1.0, so none of the (scene)linear or pure gamma logic and blend modes properly apply anyway. But undocumented changes from one version to another is a "feature" one seems to have to live with.

The maximum floating point values of ACEScc and cct are 1.4679964.

TobiJitsu wrote:The maximum floating point values of ACEScc and cct are 1.4679964.

Theoretical max yes, but in practice such high values are almost never seen and they will be clipped in anything but float storage. Log curve value 1.0 is somewhere in hundreds when linearized which no camera is able to produce anyway (dynamic range from mid gray is way higher than currently possible).

Hendrik Proosa wrote:

TobiJitsu wrote:The maximum floating point values of ACEScc and cct are 1.4679964.

Theoretical max yes, but in practice such high values are almost never seen and they will be clipped in anything but float storage. Log curve value 1.0 is somewhere in hundreds when linearized which no camera is able to produce anyway (dynamic range from mid grey is way higher than currently possible).

There is nothing theoretical about it. Although cameras do not produce dynamic range that exceeds ACES design limits at the moment, you regularly get out of gamut values when going from an OCF log format like Alexa LogC AWG to ACEScct AP1 - and with applying grading that even can get amplified, which is why the AMPAS does strongly recommend to use ACEScc/cct only as a working colour space inside a colour pipeline processing in 32bit floating point and not using it for storage or image transfer.

However, I realize this is all beside the point - I also would very much welcome a blend mode that allows for applying scanned film grain inside ACES Projects

DR16.1b3 still no fix

DR16.1 out of beta and still no fix!!! Would someone from BMD please acknowledge the issue?

I have also experienced this bug since Resolve 15.3! It's not just overlay any composite mode using a layer mixer node does not work as expected!! I'm on the newly released 16.1 and the bug is still here

DR16.1.1 is out and still no fix!!! So disappointing this issue is being totally ignored by BMD

We have someone looking into this but its not trivial so taking a little while.

Peter Chamberlain wrote:We have someone looking into this but its not trivial so taking a little while.

Great! Thank you so much!

Peter Chamberlain wrote:We have someone looking into this but its not trivial so taking a little while.

Thx

Peter Chamberlain wrote:We have someone looking into this but its not trivial so taking a little while.

Great news.
ACES is kinda flawed since version 15.2.3 on my machines.

I just wrote a 3 page explanation of how this works one some machines and some not and on friends and production house having different experience. Then I clicked Submit it and I was logged out and the post is lost...

So here is the short version.

Was on 15.3 Lite on my main computer long time ago, having this setup for Grain:

This is made in Clip Panel so to add this to entire Video

Add 2x Layer Nodes (2x ALT+L)
Remove Feed from source to lower Node
Add Film Grain Matte to Lower Node
Add Alpha Channel from Matte to Lower Node.
Set Composite Mode to Linear Light
Set Key to about 0.300 (to your taste)

This used to yield exceptional results, with a very natural grain that merged with the Highlights and the Shadows.

I upgraded to 15.3 Studio and now exact same setup gives extreme Gamma shift. No matter what Composite mode, there is always either a slight or heavy gamma shift.

So I revert back to 15.3 Lite. The Gamma Shift stays, the Layer Mixer is broken.

Also when Disabling all nodes (ALT+D) in this setup after upgrading to Studio the image goes black, this did not use to be prior to the broken Layer Mixer.

Same for Version Studio 16+.

Been abroad for 3 Weeks.

I brought my crappy Laptop that has been running version 15.3 LITE. Strangely enough on this Laptop the Linear Light Composite Mode never worked, it Gamma Shifted like the Studio version does. I have been messing about with some grades and what not, playing around with Cinegrain and I stumbled upon a Workaround.

Workaround:

Add 2x Layer Nodes (2x ALT+L)
Remove Feed from source to lower Node
Add Film Grain Matte to Lower Node
Add Alpha Channel from Matte to Lower Node.
Switch the Feed from the two Nodes to the Layer Mixer, so Lower Node goes in Higher Input and Vice Versa.
Set Composite Mode to Multiply

And Voila! You get the beautiful grain with no Gamma Shift, merged beautifully with Highlights and Shadows. Also Disabling all Nodes now does not make the screen black (ALT+D). Only Caveat was that you could not Key out the grain this way, but I have full collection of Cinegrain so I have been able to find a suitable stock everytime.

Happy Happy days, so today I upgrade Resolve on this Laptop to Studio 16.12 and now this Workaround does not work anymore, instead it clips highlights like crazy. Disabling Nodes does not make the screen black though.

I feel these issue are very random at best, like some code being changed in one part messing up another? What is being changed under the hood causing this?

Edit: So this was the short version and I forgot to add that the issue is that the LAYER MIXER is obviously broken.

EDIT II: I just noticed now that this is in the BETA section, I am not running Beta but the issue is still there.

See also: L*a*b Color space issue: viewtopic.php?f=21&t=88475&p=548029&hilit=LAB+color+space#p548029

DR16.2 issue is still present

yup still broken.

for months now I have been unable to use any grain stock in resolve.

I reverted back to Studio 15.3.1 after playing around with Studio 16.2 the Color Managed mode is worse than before, now setting Input Color Space - Linear (RAW FOOTAGE) to OUTPUT Gamma and Color Space of any variant will clip all data at around 70 IRE. Effectively making Color Managed workflow useless now.

I keep feeling blackmagic is pushing us to just adobt ACES workflow, but I find it unnecessary to go with the ACES workflow when I dont have a bunch of cameras to match up and adding a lot of unnecessary steps when I just want to grade directly on the log and not rely on IDT and LUT's.



Set Layer Mixer to Multiply.

Only caveat is that you can not adjust the key of the matte. But personally, with a collection of Cinegrain and other grain packs I've managed to find the right stock for the right footage, it does exclude some of the heftier grain stocks in lots of cases because they are too grungy.

Here's to 16.3!

Peter Chamberlain wrote:We have someone looking into this but its not trivial so taking a little while.

Hi, Peter,

Please could you give us an update on this issue? Will it ever be fixed or is there another way of adding grain in ACES that you could recommend?

Thanks!

Is the workaround presented by John Wiley Smith above not working for you?

DR17 is looking great, but I was shocked to see the issue is still there!

No workaround I've tried was of any use. I still have to export to PR4444 and apply grain in REC!. And just to think that 15 months ago and 2 releases earlier things were fine...