Adobe, Enhance Details, DNG Files, and Film Simulations

Way back on episode 40 of the FujiLove podcast—which, if you are reading this site and you like podcasts, you should probably be subscribed to—Jens and Billy Luong of FujiGuy fame had Product Manager Sharad Mangalick from Adobe on to talk about updates to Lightroom and ACR, and specifically the new feature “Enhance Details”.

Enhance Details

For those who don’t yet know, Adobe’s own dialogue box says:

“Enhance Details uses machine learning1 to improve details and reduce artifacts in most RAW files. The enhanced results will be saved as a DNG file.

It’s a computationally intensive process that performs best with a fast GPU. The estimate Adobe gives me on my Late 2015 5K iMac for a single image is 5 seconds, and that seems about right. What’s really unfortunate is the file size of the resulting enhanced DNG version of a 24.6 MB compressed RAF out of an X-H1 is 117.7 MB.

The results out of this feature would have to be pretty spectacular for me to consider taking on that kind of additional data.2 I’m seeing some decent results in my own limited testing, but weirdly, the Enhance Details dialogue preview totally betrays what the feature is actually capable of in some cases. I had one image that looked significantly worse in the dialogue box preview—more false detail, crunchy texture where there shouldn’t be—and I was ready to call Adobe out on it. After I actually rendered the image though, it did look a fair bit better than the default rendering.

It’s fantastic that Adobe are putting in the time and effort to support X-Trans—they could easily have made this a Bayer-only feature since they obviously support many more Bayer cameras than they do X-Trans. I’m looking forward to the day this feature is part of Lightroom proper, vs. behaving more like a plug-in, but one comment from Sharad on the podcast really jumped out at me:

“Fuji’s proprietary RAW file, the RAF, it doesn’t have an openly documented specification that Adobe can use to add the additional information.”

This is the result of Adobe having to create a separate (huge) DNG files. An obvious question is why doesn’t Fuji open up the gates enough to allow Adobe to write this additional information to the RAF/a sidecar file so we don’t need to go through this cumbersome workflow? I’m sure Billy has asked Japan this question, and either his hands are tied or there’s a really good reason that he agrees with.3 Either way, I wish we got an inkling as to what the issue is there.

Another obvious question is why are other applications able to generate better results without having to create a separate file? What does Phase One know that Adobe doesn’t? Adobe talk about having to balance performance with resolution, but Lightroom isn’t exactly blowing away the competition when it comes to performance with RAF.

For the time being, unless you’re all in on Adobe and a Creative Cloud subscription maxes out your budget allotment for photo editing (which is perfectly reasonable), a strong case can still be made for better, more specialized tools like Iridient Developer for those really important or really challenging images.

Film Simulation Modes

In the back half of the interview, the trio speak about Film Simulation Modes and how Adobe works with Fujifilm to ensure they have the same understanding of how each Film Simulation Mode should affect an image. This has been the official story for years now, and back when it was first told, Adobe’s interpretation of Fuji’s Film Simulation Modes were nowhere close to what we saw out of camera. To my eye, they’ve gotten better, but my recent experimentation with Capture One 12 suggests Phase One is quite a bit closer. Indeed, Phase One’s interpretations of Fuji’s Film Simulation Modes so far has been strikingly good. I’m on the cusp of switching to Capture One for much of my X-Trans processing needs for a few reasons, not needing to create whole separate files to get better detail out of my images and Film Simulation Modes are two big ones.

  1. What doesn’t use “machine learning” these days? Anyone else already tired of that? ↩︎
  2. In the early days of DNG, I remember reading about photographers converting to DNG wholesale with some going so far as to then delete the original RAW files. I’m awfully glad I never considered this kind of asset management. ↩︎
  3. Billy also does a remarkable job navigating these waters. He’s in a tough spot having to balance transparency and trade secrets, but he comes across as really quite genuine. Having spoken with Billy myself quite a few times in the past, I can say he truly does take user feedback to heart, and when he says he’ll take something back to Japan, he means it.

    I’m really digging this addition to FujiLove. It can come across as an advertisement at times, but it’s counterbalanced with honest upgrade recommendations to listeners who could perhaps skip a generation. ↩︎