Computing Life

TL;DR: I got my hands on the DJI Pocket 3 today. After using it for a while, I found its portability and ease of operation to be truly impressive, yet the image quality is quite disappointing, not matching up to the iPhone's standards.

The contradiction between the size and weight of photography equipment, ease of use, and image quality has always been irreconcilable. As I often need to film children, sometimes at home, sometimes at parks or zoos, I've experimented with four different types of devices.

• First, the cinematic-grade cameras used in Hollywood, like the RED Raptor. The image quality of these cameras is top-notch. Whether it's the feel of color grading in post-production or the detail visible on a TV screen when viewed up close, it's flawless. However, its biggest issue arises when it’s equipped with specialized cinema lenses and stabilizers, making it quite heavy—easily weighing around 10 to 20 kilograms. It's manageable to use at home for about 20 to 30 minutes, but once I need to go outside, not to mention the hassle of assembly and balancing, just carrying it for two hours while following kids is physically challenging, so I generally don’t bring this setup when going out.
• Second, consumer-level cameras with video recording capabilities, like the Canon R5. The image quality of this equipment isn't really a bottleneck in everyday use. Its differences from a cinema camera only become apparent in extreme conditions like strong sunlight mixed with shadows. Compared to cinema cameras, it's much easier to operate. The greatest advantage is that it starts shooting within a second or two of turning it on, unlike cinema cameras that require about a minute to boot up. But I rarely use this setup, mainly because there is another type of equipment available.
• Third, smartphones. Before the iPhone 15, the difference in image quality between smartphones and cameras was quite noticeable. Images had a smeared look, and colors were overly rich. Simply put, the heavy processing of computational photography made it impossible to salvage the quality in post-production. But the iPhone 15 has completely changed that; it can record directly in ProRes, offering RAW-level image quality, and its color science is also very well executed. When color grading in post-production software, it's hard to believe that the footage comes from a phone. Additionally, smartphones have the huge advantage of multiple lenses, which greatly facilitate shooting. For instance, you can switch to telephoto or ultra-wide with the press of a button in the middle of shooting, something that cameras cannot do. However, its dynamic range and other quality metrics are still inferior to those of a proper camera, which becomes apparent in many scenarios. But it's just so portable. With an Osmo Mobile as a mobile stabilizer, it can be easily slipped into a pocket, unlike a camera that requires carrying a bag. Therefore, I find it hits a very good sweet spot in terms of image quality and ease of use. Yet, the iPhone also has its drawbacks, mainly that to achieve reliable, stable footage, it still requires a stabilizer. Fitting it onto a stabilizer can be quite cumbersome, plus it means carrying an extra item. Also, because I use third-party shooting apps, and the Action Button is occupied by GPT, starting the shooting app always takes some time. Overall, there is significant room for improvement in both image quality and convenience.
• Fourth, action cameras. I've previously used Insta360's panoramic and action cameras, and while they are incredibly convenient to operate, the image quality is very poor, immediately recognizable as being on par with or worse than smartphone-level. Recently, I've seen many people endorsing the DJI Pocket 3, so I also rented one. Once in hand, the control was truly impressive; just by flipping the screen, you can start shooting within a second. Plus, it has built-in stabilization effects, eliminating the need to assemble a stabilizer and open a shooting app, greatly improving the overall experience. Another very important detail is that sometimes, for instance when you're filming kids and suddenly one starts crying, you need to quickly put down your equipment and go to comfort them. Whether it's a smartphone with a stabilizer or a camera with a stabilizer, it generally takes about 5-6 seconds to pack it up, then you can go and console the child. But with the Pocket 3, you can really start doing something else within a second. This extremely rapid and convenient contact switching was a significant advantage I hadn’t anticipated. Moreover, its specifications look very professional, with a 1-inch sensor, D-Log, and 10-bit HEVC.

However, when I actually began post-production, I encountered a major obstacle: the image quality was actually on par with the Insta360, indistinguishable as having a 1-inch sensor. Specifically, its color grading felt very similar to a smartphone, with greasy colors that became unnatural with just a slight tweak of the curves. This left me quite puzzled. Was it the quality of the sensor, or was it that the ISP pipeline overly processed the image, or was it an issue with the encoding? So, I conducted a simple experiment: first, I took a photo of a scene in camera mode using RAW format, and then I recorded the same scene in video mode using HEVC encoding. The comparison is shown in the image below:

The difference seems quite obvious to me. The redness on the little faces appears striated in the RAW photo, but in the video, it’s blurred into a blob. Additionally, the video’s colors and brightness are somewhat bizarre. Here, I used the standard REC.709 mode without recording a LOG curve or applying a LOG LUT.

I’ve conducted similar experiments with the Insta360 and obtained very similar results. So, I now tend to think that currently, these action cameras have a major shortcoming in video encoding. Even with excellent processors and image processing pipelines, the images are severely compressed by these encoders, leaving very little room for post-production and resulting in poor image quality.

In summary, I am quite disappointed with this DJI product, although it indeed resolves some usability pain points of smartphones. To date, its image quality has not come close to matching the iPhone 15’s ProRes footage. The gap is significant. However, I believe this isn't an insurmountable technical issue, and perhaps in the next few years, we might see action cameras employing better encoders, maybe even internally recording ProRes like the iPhone.

So, my current filming equipment setup remains two-fold: the cinema camera and the iPhone 15. Without physical or mobility constraints, I use the cinema camera for filming, but if I'm going out, especially for several hours with kids, I stick to the Osmo Mobile with the iPhone 15. I am very satisfied with both setups.