Choo! Choo! Freeze64!

Freeze64 issue 68, featuring an illustration of a train and riders on horses

Well, chug along and check this out! Freeze64 Issue 68 steamed its way into my mailbox, featuring a review of a train robbers game I never knew existed. Plus, there’s another captivating instalment in Andrew Braybrook’s Uridium “making of” series. I always enjoy diving into the technical details of retro development. You’re constrained so much, it’s a challenge fitting a game into such an environment.

Impossible Mission III looks great too. I wish I was able to get further in the original game. Maybe this one will be easier. C64 games are hard as nails, aren’t they? Quite unforgiving compared to (many of) today’s games.

Lossy DNG isn’t so bad

Have you ever converted a RAW file to lossy DNG and noticed how much smaller it was? You can make the RAW file 80-90% smaller! I always thought of lossy DNG files as JPEG files saved at a compression level of 10 or 100%. In other words, with barely anything stripped, but it’s more interesting than that.

I discovered that a panorama I made in Lightroom was a lossy DNG. After I made it, I tried converting it to lossy DNG, and the new file was the same size as the old one. So, panoramas are lossy DNG. They used to be huge files, but, in Lightroom 13.0 that changed. Adobe started using Jpeg XL instead of Jpeg to store data in lossy DNG files. Even though the files are smaller, they are higher quality than older panoramas! It’s the same with HDR and Denoise. Remember when Denoise was introduced and the files it made were gigantic?

HDR images created by Lightroom are also now saved using the same lossy DNG format. An HDR image I created in 2017 on my 24MP Sony A7III is 77MB, but using Lightroom 13.3.1 that file is only 11.9MB. When I convert the image manually to lossy DNG with an embedded medium Jpeg and fastload data, the file reduces further to 8.1MB and looks practically the same at 100%.

I haven’t noticed any major problems fixing highlights or shadows in panoramas, so lossy DNGs are pretty good, but not without their own drawbacks, which I’ll get to later.

On the Adobe forums someone noticed this last year, and received a few interesting replies:

The compression method for derived DNGs has been changed from JPEG to JPEG-XL, which provides a smaller data footprint without loss of quality.

Creating a merged DNG such as a panorama is already a “lossy” process because the merged pixel data has been demosaiced, aligned, and blended from the original photos. Using JPEG XL compression makes a much smaller visual change.

Do not look at the words ‘lossy’ and ‘lossless’, look at your image and see if there is any visible effect of this new compression method. Recreate an older panorama and compare the old massive DNG with the new, much smaller DNG. Do you have any reason to be concerned?

Before you convert all your RAW files to lossy DNG, be aware that it will affect how Lightroom treats your file in ways you mightn’t think of. Years ago, I noticed that the Transform tool worked differently on lossy DNG files. It straightened walls slightly differently, not that it looked wrong, just different. You also can’t feed DNG files to Topaz Photo AI (Files->Plugins Extra->Process with Topaz Photo AI) or Lightroom Denoise, so use those before making your panorama or making the file a lossy DNG.

This is an example of a simple image that transforms differently when it was converted to a lossy DNG file. Lens correction has been applied to both images, and they are identical then. However, when “auto” transform is applied, the lossy DNG is modified differently. This bug has been there for years. If you want to compare yourself, here is the before image, and here’s the after image.

Greg Benz has published a post on the subject too and came up with the idea of exporting images to lossy DNG “to the same folder” and adding them back to the catalogue, making it easier to compare before and after images before you delete files. He points out some other problems I never noticed with lossy DNGs, such as haloing and colour tones, and other unspecified issues with older RAW images imported before 2012.

Is it worth converting your RAW files to lossy DNG? You’ll make savings of up to 90% and that’s very tempting. A 66MB RAW file may become a 9MB lossy DNG file. If you don’t like deleting, but you’re sure you’re not going to use an image, then it may be an option for you. On the other hand, use with caution on photos you want to edit and publish.

A comment on this video suggests loading the original photo and the lossy version into Photoshop as layers and use the “difference” blend more on the top layer. You’ll notice tiny changes between the images. Since I have deleted the original RAW file of the photo above, here’s one of the lighthouse at Ballycotton.

Here’s what the difference looks like in Photoshop.

There are obvious differences when you look at it like this, but I still think it’s worth considering. Today’s photo on my photoblog, The Rocks of Ballycotton, is a lossy DNG file for example.

And finally, one thing to consider is that third-party software support for lossy DNG files might not be what you expect. I haven’t tested any of the open-source software out there, so I’m not sure how good it is. A comment on that video I linked to above says RawTherapee handled them, and that was 6 years ago, so I’m sure the situation has improved since then.

While in the process of researching this, it has been so nice to find actual open, public, online forums and blogs talking about this, and not just on Reddit either. I bet there must be plenty of conversation about it hidden in Facebook groups too. Hidden from prying eyes.