When was the last time you opened your panoramic camera? Several long-time users of panoramic cameras told me that these devices tend to “gather dust.” They had bought them for cool sports videos, but after receiving the camera, they realized the real issue wasn’t the camera itself—it was that they didn’t have enough time to go skiing, diving, or mountain climbing. After a few uses, the camera ended up in a drawer.
However, more friends who bought a panoramic camera in the past year gave a completely different answer. Their usage scenarios are no longer limited to outdoor sports; parenting, travel, and road trips are now their more mainstream use cases.
Looking at the overall imaging market, compared with the gimbal cameras that exploded in popularity this year, panoramic cameras are growing more slowly. The core reason is that the goals and task complexities of the two categories differ: the primary task for gimbal cameras is “shooting good portraits,” and the way to achieve that is relatively clear and constrained; panoramic cameras, on the other hand, aim to take over the entire chain from pressing the record button to delivering the final footage, achieving “unobtrusive recording.” The ceiling for this is much higher, so the progress naturally moves more slowly.
Over the past few weeks, I got early hands-on experience with Insta360’s brand‑new panoramic camera, the X6, and took the opportunity to update my understanding of the product and technology in this category.
---
The biggest problem with panoramic cameras is no longer “shooting.” Poor image quality was the biggest stereotype and negative label for panoramic cameras over the past decade. Even with 8K, because panoramic cameras capture more information, the pixels allocated per unit area are fewer than on traditional devices, resulting in lower resolution. Hardware iteration is a primary way to improve image quality. On the X6, Insta360 has installed a 1/1.1‑inch square sensor that is better suited to panoramic shooting. It can be said that the image quality issue has been largely resolved for now and is no longer the core “pain point” that drives users away.
But moving from “unobtrusive shooting” to “unobtrusive output,” the post‑production editing of panoramic footage remains the last‑mile pain point for users. Over many years, Insta360 has sought to lower the barrier to this process, building a set of “smart editing” experiences on the software side. However, if you have actually used a panoramic camera, you will find that the previous experience was still far from “unobtrusive.” The biggest flaw was that it still required repeated, fine‑grained “manual intervention” throughout the process. After shooting, you had to transfer the footage from the camera to the mobile app, select from multiple clips, and then use the phone’s or cloud’s computing power to produce the final video. But in many real‑life scenarios, I don’t have that much time for these operations.
For example, after a full day of hiking or skiing with a panoramic camera, you have to drive dozens of kilometers back home, and then go to work the next day… Over time, the footage stored in the camera or on the SD card piles up like unopened books on a desk—buying counts as reading, shooting counts as editing—and the more there is, the less willing you are to edit.
The new feature I was most satisfied with on the X6, “AI Director,” addresses this problem. Its logic is very simple: no manual uploading or selection of clips, no internet connection required. The camera automatically analyzes the previous day’s (natural day) footage each night, automatically completes the editing, and then pushes the result to your phone. All the work runs locally on the device. You wake up to a finished vlog from the day before. The whole process does not require connecting to a phone or selecting clips; the only thing you need to do is ensure the panoramic camera is plugged into a power source. Then, the “smart editing” experience evolves from a question‑and‑answer chatbot into an agent that works autonomously—much like raising a lobster on a Mac Mini while it’s plugged in.
Of course, besides the “AI Director,” which offers a local L4‑level experience, users can also use the existing “One‑Tap Edit” function in the mobile app or choose the “Cloud Moment Pro” mode that calls upon cloud models.
But here is another question: with the explosion of information in panoramic videos, how does “AI Director” know what matters most? On this front, Insta360 has paired the X6 with a nifty little accessory: the Head‑Tracking Module. This module was originally launched with the gimbal camera Luna series, and its core logic is “wherever the head turns, the lens follows.” As a panoramic camera, the X6 has no physical gimbal; the user’s head turns and pauses become the “attention input” to the “AI Director.” The same accessory transforms from a physical‑world “remote control” into a model‑world “prompt”—it can manage both capture and intention.
On the consumption side after content production, panoramic videos have long had a natural experiential limitation: you can turn your head to look around and change the viewing angle, but you cannot walk around within the scene—meaning you can change the “perspective” but not move the “viewpoint.” Released alongside the X6 is a feature in the app called “3D Time Capsule.” Simply put, this feature uses 3D Gaussian Splatting technology to “reconstruct” the captured scene. In this generated scene, users can move freely and more immersively return to the shooting location from different positions. I remember often seeing students use the “Bullet Time” feature of panoramic cameras to capture graduation moments; with scene reconstruction, what remains in the future will not only be classmates but also the classroom, dormitory, and playground that hold shared memories.
It should be noted that at this stage, 3D Gaussian Splatting is mainly geared toward static spaces, so this feature is still in an early‑adopter phase and will likely not be used as frequently as “AI Director.” Yet the overall direction is clear: on the X6, Insta360 has clearly placed the iterative focus of product experience on the entire intelligent workflow after the shutter is pressed.
---
Teaching Panoramic Cameras to Have Aesthetic Sense
From panoramic shooting that “liberates framing” to smart editing that “closes the loop” on tasks, panoramic cameras have always been striving to cross the chasm from usable to truly good. The next step is to add aesthetic judgment to intelligence.
In practice, the goal of “improving aesthetics” can be broken down into two dimensions. First, the system needs to understand the different baseline standards across various scenarios. Compared with shooting good portraits, panoramic imaging faces more numerous and more complex scenes. A skiing video and a parent‑child record have completely different standards in terms of rhythm, single‑shot duration, negative space, and the choice of highlight moments. No single set of universal rules can cover everything; it must be trained scenario by scenario. According to GeekPark’s understanding, Insta360 has conducted over 10,000 hours of training across 30+ scene types, building a separate aesthetic standard for each category using existing footage.
Second, within each specific niche scenario, the system attempts to understand the individual aesthetic preferences of different users. In Insta360’s view, each user’s preferred video style is as unique as their TikTok or Xiaohongshu feed. Therefore, on top of data pre‑training, Insta360 also tries to infer user preferences from their behavior. For example, when faced with different versions of a finished video, what the user changes and which version they finally export can become learning material, enabling the user to receive final footage that better matches their own aesthetic standards in the future.
For the model to become smarter, the chip must be more capable. At this stage, the X6 is packed with three chips, including a 4nm SoC found in flagship phones and two dedicated imaging chips. Yet even so, faced with the simultaneous computational demands of spherical stitching, high‑pixel throughput, and on‑device inference, it can only achieve “instant capture” rather than “editing while shooting.” Limitations in image quality also stem from hardware—panoramic technology itself does not conflict with high image quality; the current contradiction lies in the fact that volume, power consumption, and heat dissipation area are all fixed, so trade‑offs must be made. To squeeze out more computing power and higher pixel counts within a fixed physical space, the only paths are more advanced process nodes for chips and higher‑density sensors—neither of which has a ready‑made supply chain; the company must forge its own way.
---
The Second Decade of Panoramic Imaging
Over the past ten years, the panoramic industry has benefited from two waves of industrial “spillover.” The first came from the miniaturization of components driven by smartphones—sensors, batteries, and SoCs became smaller and cheaper, providing the industrial foundation for young startup teams like Insta360 to turn panoramic cameras into consumer hardware products. The second was the talent spillover from the first wave of AI focused on image recognition. Today’s smart editing capabilities largely evolved from the algorithms, talent, and engineering experience accumulated in the AI 1.0 era.
But as hardware—chips, sensors, and the like—gradually hits its limits, in the second decade of the panoramic industry, there will be fewer external forces to borrow from, and the company will inevitably have to create more of its own waves. Continuing to pioneer in a category that is still defining its own end point may be both a boring and interesting endeavor.
The “boring” side is that competition becomes increasingly homogeneous, and differentiation gets harder to achieve. Once a mass‑produced device is disassembled and studied, hardware solutions can be quickly matched, and software approaches can be rapidly imitated—not only in imaging but even among the “big three” AI models, the same situation applies.
But the “interesting” side is that when you shift your gaze from the specification competition of a single generation to the strategic decisions and bets made at each stage of the category’s evolution—for example, shifting the focus early from shooting to post‑production, or treating user attention as a new kind of signal—these directional judgments are what truly create distance between players.
Compared with its first decade, Insta360 has significantly adjusted its positioning of panoramic cameras. Ten years ago, the primary selling point of a panoramic camera was the “cool” shots that no other device could produce; ten years later, it aims to bring “unobtrusive recording” into more scenarios and deliver better finished footage. It can be said that Insta360—a company that started with panoramic cameras and still has “selling cameras” as its core business—is actively working to make users “forget the camera” and forget the existence of traditional editing workflows. The center of gravity is shifting from the lens itself to the algorithms and models behind it. With every step forward in computing power, the camera becomes more “transparent.”
