BM releases Davinci Resolve 21 Public Beta 1

Any testers finding less “chatter” on AI windows? Every time I try to use object selection on actual objects I give up and draw spline based windows, track and tween it after. This is on 3x HD res RED footage shot with a multi axis arm, so detail and predictability are there.

Definitely like render in place as I share projects with a client and they can miss things, will render in place work with good old fashioned windows too?
 
I have tried and am saying it is not a proper linear signal from sensor, highlights are clipped, I see in other RAW debayering software there is still information in these highlights. It is also just an inverse transform. No matter if I use color managed or unmanaged project settings.

I've seen the exact same issue in all my tests of the initial v21 beta. All values over 1.0 in linear are being clipped in raw photo formats. And .DNG ProRAW files from the iPhone show an additional problem where it appears that the Profile Gain Table Map is being baked in when it shouldn't be.

I assume BMD will fix these issues in updated beta versions.
 
I've seen the exact same issue in all my tests of the initial v21 beta. All values over 1.0 in linear are being clipped in raw photo formats.
How did you test this, and which formats? When I decode .ARW or .NEF to Linear there are plenty of pixels above 1.0 although not nearly as high as shooting log video. I wonder if this is an inherit limitation of the stills raw format and the way data is stored to it or something else. I don't have a direct comparison of a camera shooting both to eliminate sensor limitations.

Most of the 'expected' dynamic range people are used to seeing in photo editing apps is from highlight reconstruction algorithms.
 
I've tested .DNG ProRAW from an iPhone 16 Pro that I have, and some Canon .cr3 that I was able to find online.

I was wrong to write that data is being clipped. What I should have written is that the way raw stills are behaving in Resolve gives that appearance.
Recorded data probably isn't actually being clipped, but something about how it's being treated makes it appear that way in RCM, ACES, and in any manual log intermediate setup that isn't direct decode into a display space. The Canon .cr3 files appear in the raw tab with a "highlight recovery" option, and that does deliver more in the highlights when checked on in some cases. However, where those highlights land in any log/linear pipeline setup still seems off. As noted, it's lower than what would be expected from a log video recording of the same scene.

I agree that direct comparison of raw video and raw stills from the same camera would be illuminating. I'm hoping to borrow a Canon R5C from a friend sometime soon.
 
Last edited:
The team at Video Village have encountered the same issues with the v21 beta photo support:

"Resolve's RAW photo support is limited and has color-accuracy problems. Hopefully this will improve as it moves out of beta. For now, Blackmagic claims support for Sony, Fuji, Nikon, Canon, and iPhone RAW, but many of these images render with subtle to dramatic color and tonal inaccuracies in scene-referred workflows, especially Fuji and iPhone ProRAW. Even when color is accurate, images often come in quite dark."

https://videovillage.com/blog/2026/05/12/filmic-photography-resolve-21
 
Note that VideoVillage is giving away RawZone to anybody with a valid account on their website:

Would be interested what is behind it.

We are at the moment using RawToACES for few stop-motion animated projects.
 
Back
Top