I recently wrote about how to move the excellent photos from the Google Pixel phone Camera app into a desktop Lightroom workflow. I was pleased that it’s easy to tell the camera to generate DNG “RAW” files and include them in the process. But apparently, the camera’s JPGs are better and more useful than the DNGs. That’s weird.
Here’s a pair of pictures to illustrate. This morning, the cat found a sunny corner of the back porch and was squirming around out of pure joy.
The DNG is above, the JPG below. Of course they’re both JPGs now in the blog, but both are straight out of the camera, resized and JPGized by Lightroom with no sharpening or anything. It’s not that dramatic here, but flipping back and forth in Lightroom, the difference isn’t subtle. The JPG has had some lens correction, the blown-out highlights have been recovered, there’s been a bit of sharpening (look at the cat’s belly hair and the broom bristles), and the color’s been tweaked — that watering-can is dead-on in the JPG but has extra yellow in the DNG.
In this CNET piece, Marc Levoy, who invented the term “computational photography”, says “The JPEGs from the Pixel camera may actually be more detailed than the DNGs in some cases” and yeah, no kidding. In fact — and this puzzles me — the JPG is 4032x3024 in pixels while the DNG is 4016x3008, which is to say it’s 112K bigger. But I don’t think that’s what Levoy meant.
Also: “Our philosophy with raw is that there should be zero compromise,” Levoy said. “We run Super Res Zoom and HDR+ on these raw files. There is an incredible amount of dynamic range.” That doesn’t match my experience, but then he was talking about the Pixel 3 and I still have a 2. Also, Stephen Shankland said: “I rather like Google’s computational DNGs from Pixel 3. They HDR-ize the raw input to create the DNG. It’s not perfect but I find it darned useful. (I also generally like the Pixel camera app’s JPEGs, though they can look overprocessed to my eye.)” So I look forward to giving Pixel 3 (or 4) DNGs a try.
Why do you want DNG anyhow? · Photographers like RAW versions of photos because they’re more editable. One of the most common editing modes is rescuing lost data from highlights or dark areas that look blown-out or dimmed-out in the original — you hear people saying that a good camera RAW is “deep”, and that’s certainly true of the files from the Fujifilm X-cameras.
Consider these three pictures. Once again, the first is the DNG, the second the JPG, and in the third, I decided to see if I could recover image in the dark area behind the hydrangea blossoms. (Not a thing I’d normally do on this shot, I like the dark framing.)
I think the results speak for themselves. There’s a lot of useful data in this JPG.
Which raises one more question: By aggressively digging in with Lightroom, could I replicate what the Google camera software did, or maybe even improve on it? So I tried that, and there was progress but at no point did I think I was really replicating that tasty JPG, and I got bored trying.
So for now I think I’m going to turn off the DNG capture on the camera app. Sshhh, don’t tell any Real Photographers.
Comment feed for ongoing:
From: Jon Ellis (Jul 28 2019, at 03:49)
in my experience (with iPhone) you can reproduce the "look" by:
- increasing local contrast (clarity) +15
- increasing the vibrance by +10
- increasing the sharpening, to around ~50 - 75
- hitting J and bumping the shadow area / pulling the highlight and whites
however, if you do that to everything you process it starts to look hyperreal and unnatural. that might well be a personal preference / aesthetics thing...
From: Yaron (Aug 24 2019, at 22:41)
TL;DR - With 16 bits/channel in the DNGs versus 8 bits/channel in the JPEGS, shooting in DNG will give you better pictures when you revisit them in the future. But I agree, it makes no difference today.
Long Winded Version -
Hi Tim it's been... a REALLY long time but I randomly ran across your blog while I was looking for people with the same issue you raised, e.g. is it worth taking DNGs on the Pixels (I have a Pixel 3A) when the JPEGs are so amazing?
Asking the question forced me (rubber duck debugging style) to get an answer.
I think the short answer is - the DNGs are 16 bit per channel and the JPEG are 8 bit for channel  So as screens get better the DNGs will look better and better, the JPEGs won't. And as the software gets better it will be able to do more wonderful things with the DNGs than the JPEGs.
But for today on my 8 bit monitor the JPEGS look just about as good as anything my limited skills let me do in Lightroom with the DNG . So if I only care about today, JPEG is the way to go. But in the future lightroom will get better and the data in those DNGs will give Lightroom way more space than it will have to improve things tomorrow than it can with the limited JPEGs.
So today I shoot with both.
 There are extensions to JPEG that allow it to encode greater than 8 bits per channel but I checked some of my Pixel 3a images using magick and they don't appear to be using those features.
 Here is what I do
1. Hit tone/Auto
2. Bump Vibrance up by around 40
3. Bump up Saturation by 20 or so points
4. Bump up Sharpening/Amount by 30 or so
5. Texture up by 50 or so
6. Clarity up by 20 or so
7. Enabling Profile Corrections to Google Pixel 3 Rear also helps a tiny bit but it's only noticeable if you have the photos side by side.