ACES on set viewing LUTs (Blackmagic)


(Thomas Wilshaw) #1

Hi,

I’m wanting to create a viewing LUT for on set monitors but I’m having some problems. Normally I would concatenate an IDT, RRT and REC.709 ODT using the OCIO tools (is this an acceptable way of doing it?). However in this case I’m working with The Blackmagic URSA which has no published IDT. What I’ve tried is running the OCIO LUT image through resolve with the URSA IDT and the correct RRT/ODT and then pulling a LUT out of that with OCIO again.

It sort of worked but there’s a slight tone shift in the image, almost like an incorrect white balance, when I compare applying the LUT to an image and using the proper workflow on the same image.

I’ve only used a 64x64x64 cube so could that be the issue? Also the footage I’m testing it on is raw so are there any extra steps going on there that the LUT image wont have taken into account?

I did get in touch with Blackmagic but this was there response:

The best way to generate a LUT that represents the ACES transform would be to grade one of your standard clips to, match the clip created using the ACES colour space workflow. You can then generate a LUT from this and import it to your monitors.

Which seems a little hacky to me, any advice?

Thanks


(Willian Aleman) #2

Hello Thomas,

This might not answer your question directly, but it’s something to take into consideration. As far as I know the only LUT box capable of higher 3D LUT size, like the one you are creating is the eeColor 3D LUT Box, (65x65x65) with the limitation that it only supports HDMI I/O, and the same is true for cameras and reference monitors internally supporting 3D LUT.

It would be interesting to know why are you using 64x64x64 3D LUT for on-set viewing, when there is so little support for on-set purpose. What LUT box are you using that support 3D LUT of this size?


(Thomas Wilshaw) #3

Hi, thanks for the reply.

Sorry I’ve not explained that well, I will be using a 33x33x33 LUT on set and I started with that in Resolve. I used the 64x64x64 (in Resolve) to see if it was LUT size causing the issue but it doesn’t seem to be.


(Willian Aleman) #4

Hi Thomas. Sorry for the delay. I have been on-set in remote area for the last two weeks, where there is no Internet signal until we return to the city at night.

Regarding your question about the tool you are using to create the LUT, as a DIT, I would recommend using one of the standard tools used on preproduction and on-set for color management and creating looks: LiveGrade, Prelight, FireDay, among others. Each one of them support ACES workflow.

Since DaVinci Resolve Studio doesn’t support LUT boxes and since Blackmagic Design own LUT box, HDLink Pro has no driver support since MacOS 10 later versions, working with Davinci Resolve is going to slow down the workflow on-set. Unfortunately, DaVinci >Resolve Live< is impracticable to be used on-set for live monitoring purpose, and DaVinci Resolve 14 didn’t bring any good news regarding color grading live on-set.

Regarding the shift tone: It’s probably that the shift in tone you are seeing is part of the nature in the way ACES RRT processes the images. It has its own mapped tone, its own personality. In other words, the RRT is not transparent. This has been discussed in this forum numerous times. Some colorists like it and others don’t. It’s my understanding that the issue or feature in the way ACES RRT works, (depending on what side of the effect the colorist aligns) is under discussion for the next enhancement release version of ACES. Depending on the exposure and quality of the light source footage has and the intended look of the image, sometimes the RRT works as a LUT, a miracle created specifically for that footage. With other footage and times the results can be less than desirable, making the colorist work backward to correct the undesirable RRT effect. And that inconsistency has created the disappointment or glory, hate or love for the ACES RRT.

In terms of Blackmagic Design camera IDT, DaVinci Resolve Studio, version 12.5 and 14 already supports BMD Film, (BMCC) 4K and 4.6K IDTs, in addition ARRI Alexa, Canon 1D/5D/7D/C300/C500/Panasonic V35/Sony RAW/ slog2/slog3.


(Thomas Wilshaw) #5

Hi William,

Thanks again for your help, but I’m afraid I’m probably not explaining myself clearly enough.

The final grade on the film will be done in Resolve with Aces cct.

On set, we will be very limited in terms of DIT. We’ll have someone backing up cards, but no on set grading or anything. So what we want is a LUT we can load onto monitors (a SmallHD 502 in this case) that will give a reasonable (all things considered, especially the budget) approximation of the final ACES “look” whilst on set.

What I planned to do was create a LUT that would stack the effect of the camera IDT, the RRT and a REC.709 ODT into one 3D LUT, I’ve done this in the past with other cameras with OCIO. I understand that the RRT is somewhat subjectively designed, and by and large like how it looks. I just wanted something that would show how footage will look after ACES colour management, but without a custom grade.

Unfortunately, although Blackmagic have evidently created IDTs for all their cameras, they are hidden away inside Resolve, so I can’t access them. For whatever reason, they won’t publish them openly.

To get around this, I planned to load the OCIO LUT Image into Resolve, tell it it was Ursa footage, export it through the RRT and ODT, and use it to create a LUT. This has worked up to a point, but if I then apply that LUT to footage (in Resolve) it gives a different result to running the same clip through ACES colour management in Resolve; this is the tonal shift mentioned.

I tried creating the LUT as a 64x64x64 cube in case the shift was simply due to a lack of precision in a monitor sized LUT, but got the same result. Once I’ve got the LUT worked out, I’ll make one the appropriate size for the monitor.

So, does anyone have any idea what is causing the colour shift and how I can get round it? I’m assuming it’s some quirk of Resolve’s. Alternately, are there any decent home made IDTs out there for Blackmagic’s 4.6K sensor. Or does anyone have any contacts at Blackmagic who might be willing to discuss making the IDTs public?

We start shooting on Tuesday.

Thanks


(Willian Aleman) #6

I believe that I finally get the purpose of your LUT creation.

Since there is no way to get an external IDT for the Blackmagic 4.6K sensor, the shortcut of creating a functional LUT for monitoring on-set is to create it in DaVinci Resolve.
This will mean that the exported LUT will carry the proper IDT for the camera you are using, + the grade + RRT + the ODT to be used on-set in the monitor.
I also would suggest to export the LUT you create in DaVinci with just the IDT and the grade without the RRT+ ODT if you want to preserve the look. When you bring the footage back to Resolve use that LUT and enable the RRT and ODT back in Resolve. This will also help if you need to send the LUT to the VFX department or another applications, in addition to avoid the mistake of doubling the RRT +ODT.

I’m not sure if the issue of Resolve baking ACES LUT with the RRT+ODT still persists in the later versions of the application. If this is true, you need to use Nick’s CTL scripts http://www.antlerpost.com/new_products.html to resolve the issue.

Another possibility of the shift in the tone you are having is because since you are baking the RRT and ODT in the LUT created in OCIO, you are probably forgetting to turn off the ODT in DaVinci when bringing back the LUT. This can result in doubling the RRT+ODT. This is just an assumption.

Hope this helps!


(Nick Shaw) #7

Is the footage you are recording from your URSA BMD Film ProRes or DNG raw?

This is speculation, but if you are recording raw, I believe that Resolve does not use a single IDT for DNGs from BMD’s own cameras, but rather decodes the raw directly to ACES colour space using a matrix which is dependent on the shooting white balance.

The IDT for BMF Film ProRes has no knowledge of the shooting white balance, so I expect a “one size fits all” IDT is used. Your approach of putting a CMS image through the fixed IDT + RRT + ODT to create a LUT is a suitable one to match the fixed IDT. But it will not match the variable IDT for raw. Or at least it will only do so at one particular white balance.

But I repeat, this is my speculative theory. I have no inside knowledge.

Regarding my DCTL scripts Willian refers to above, I don’t think they will be of any help to you. They can be used to create a LUT which does not include the RRT or ODT, for use in e.g. Nuke, using a method described elsewhere on this forum, but the LUT you need should have everything (IDT, RRT, & ODT) baked in, so your CMS image approach is a good one.


(Nick Shaw) #8

Actually, I believe Willian may have been referring to what I describe in this post for exporting a LUT including IDT, RRT and ODT directly from Resolve. But that is reliant on replacing Resolve’s built in IDT with one written in DCTL (such as the ARRI, RED and Sony ones I sell). But since, as you point out in your original post, the BMD Film IDT is not published, it is not possible to rewrite it in DCTL.

I did try deriving a matrix for BMD Film by running test patterns through the Colour Space operator in Resolve. But my result did not match it exactly. I can only assume that BMD apply some “secret sauce” to processing of images from their own cameras, which is not easy for third parties to replicate.


(Willian Aleman) #9

Hi Nick,

Thanks a lot for your comments.
I learned about the workflow I described in my previous writing here from your comments on Robbie Carman’s series of insights on ACES at Mixinglight.com. In this ACES series Robbie extensively describes how to setup ACES in Davinci Resolve. The insights are part of an outgoing series. I believe Mixinglight has published the ACES series as a free reading.

Thanks once again for your constant support to the ACES community


(Thomas Wilshaw) #10

Thank you both for your advice.

Nick, that seems to be it. Testing with just ProRes clips we can get a good match, bar some low res LUT imprecision, so we’ll go with the fixed IDT approach and get back to you with more details about how it works out after the shoot.

I wish Blackmagic would just publish, but there we are. I attempted to make an IDT for the pocket camera a couple of years ago by shooting a Macbeth chart under tungsten light and using the DNG metadata tags to linearise. I followed the recommended IDT procedure as well as was possible without scientific equipment, and it sort of worked, but there were some definite colour problems.

Thanks


(Thomas Wilshaw) #11

Sorry for the delay in following this up. I went with the approach outlined above (what we might call the ProRes IDT) and it worked fine for on set monitoring. I’ll do a detailed comparison of how that compares to the hard coded raw IDTs at the relevant white balances once I receive a copy of the footage. Attached is the LUT we used, both at high res and as a smaller cube for monitors. Maybe it will be useful for someone. I’ve changed the extension to pdf so it will upload, just change it back to “.cube” to use it.

URSA_46k_rec709_D65.pdf (982.7 KB)

Interestingly, at no white balance setting could we get an exact match between this LUT and DaVinci’s ACES processing of raw files, so there must be more to the “secret sauce” processing than just that. I might investigate that further when I have more footage to play with.