Ten Years of Insta360 Panoramic Cameras

Produced by | Chuangye Zui Qianxian

Author | Xiaoyun

Edited by | Lightning

Art Editor | Xing Jing

Review | Songwen

On August 13, GoPro reported its second-quarter results: revenue of $105 million, down 31% year over year, and a net loss of $51 million, with the sale review launched in May now in its "final stage." At its peak, the company held more than 90% of the action camera market; today it is waiting for a buyer.

Just the day before, Shenzhen-based Insta360 unveiled its new-generation flagship 360-degree camera, the X6, billed as "a decade in the making." At a closed-door briefing before the launch, founder Liu Jingkang was candid: the hardware has indeed hit the current ceiling of 360-degree camera technology.

But hitting the chip ceiling is not the same as hitting the market ceiling. Chips and sensors may have peaked, capping hardware performance for now; how many people a category can reach, how many scenarios it can enter, and how large it can sell are still unfolding.

Yet the new leader's battlefield is anything but calm. Last July, DJI entered the 360-degree segment with the Osmo 360. According to IDC data, in the first quarter of this year Insta360 held about 68% of global 360-degree camera shipments and DJI about 22%; by revenue, Insta360's share reached 71%. GoPro and others have become "others" in the market.

The walls are being torn down further. New players keep emerging: Photon Leap, whose CTO came from Dreame, closed an angel round worth several hundred million yuan; Looki, founded by a former Meituan executive, launched a wearable AI camera. "AI camera" seems to have become a story anyone can tell.

Against this backdrop, Insta360's 2025 annual report shows a "split" picture: revenue of 9.741 billion yuan, up 74.76% year on year; net profit attributable to shareholders of 929 million yuan, down 6.62%; R&D spending of 1.53 billion yuan, or 15.7% of revenue. The company attributes this to "increased strategic R&D investment." Liu Jingkang puts it more bluntly: the profit shaved off by selling hardware can be offset by building a new commercial loop around services derived from AI servers — essentially a hedge.

Hardware profit is being diluted while AI services are still growing. Now that cameras long ago became capable of shooting clearly and capturing everything, what more can they do for people? Insta360's answer: make the camera "invisible," and hand image quality, editing and other tasks on the device to cloud AI.

       1. From two million units to countless millions

The story begins in 2015. Liu Jingkang, born in 1991, graduated from Nanjing University with a degree in software engineering. While still a student, he built a "famous university livestreaming" product; because the video quality was poor, he turned his attention to a camera that could shoot 360-degree video with one click. One year after graduation, he founded Insta360 with his co-founders.

In 2016, Insta360's first consumer 360-degree camera, the Nano, plugged directly into the iPhone's connector, solving what was then the biggest hassle in transferring 360-degree footage. It became a star product at CES, drawing praise even from Apple co-founder Steve Wozniak.

But the hit product's momentum did not last. After sales fell back, the team reviewed user data and found that the people using 360-degree cameras most frequently were skiers, cyclists and surfers. They had no free hands to frame a shot, and the "shoot first, frame later" nature of 360-degree capture did the framing for them.

There was an even more telling clue: because the Nano needed a phone connection to work, some customers duct-taped their phones to the 360-degree camera, mounted the rig on helmets and selfie sticks, and shared the results in user groups.

From then on, Insta360 stopped forcing the same camera into more markets and instead rebuilt the hardware for each use case.

In 2017, the first 360° panoramic action camera, ONE, debuted with an invisible selfie stick and Bullet Time, breaking GoPro's dominance of the action imaging market. In 2018, ONE X brought 5.7K panoramic video and FlowState stabilization.

Insta360 EVO and Insta360 ONE R followed in successive years, while the X series kept evolving: resolution rose from 5.7K to 8K, sensors grew from 1/2.3-inch to 1/1.28-inch, and stabilization advanced from algorithms to an AI triple-chip architecture. Generation after generation of technical upgrades enabled Insta360 to leap from challenger to leader.

As of 2025, Insta360 has ranked first in global panoramic camera market share for eight consecutive years, cementing its absolute leadership in the category.

From 2015 to 2020, Insta360's cumulative sales surpassed its first 1 million units. With the launch of the X3 in 2022, shipments exceeded 1 million units in just over a year. In total, Insta360's panoramic cameras took five years to reach the first 1 million units and only two years for the second million; the interval has kept shrinking since, with the latest million taking just half a year.

As sales climbed, over the past decade the panoramic camera evolved from geek gear used to summit Mount Everest into a mainstream tool for capturing everyday family life.

Despite the rising numbers, Liu Jingkang does not believe the ceiling has been reached: "The day a panoramic camera is in everyone's hand like bottled water, tissues or a smartphone is when we should really start talking about the ceiling."

2. A data moat and a long-term vision

Qi Lu (Gavin), head of the Insta360 Research Institute, joined in 2024.

On August 12, the day the X6 launched, he told Chuangye Zui Qianxian that he left ByteDance's video generation data team after spotting a problem while working with data: "Mainstream video data is shot from fixed viewpoints, so information is inherently missing." Full-view 360-degree footage is the higher-quality data source for the AI era.

What convinced him to join was an Insta360 vision internally codenamed "Cameraman" — an intelligent system that can accurately understand user intent, shoot autonomously, and automatically edit and share.

Qi recalled to Chuangye Zui Qianxian that he first met Liu Jingkang at a teahouse in Shanghai, where they talked for more than three hours about Cameraman's business and technology. "That first meeting basically settled it."

According to Qi Lu, the Cameraman project was first proposed internally in 2018 and 2019. In 2019, Insta360 began building capabilities in three technology areas — Imaging, AI and Robotics — corresponding to the eyes, brain and body.

By 2023, Liu Jingkang had already shown Qi Lu a first-generation drone prototype and some features that had not yet entered mass production. The panoramic drone can be seen as one of the early precursors to Cameraman. "It can free the shooter's hands and enable unobtrusive shooting," Qi Lu said.

At this point, Cameraman has become clearer: its core competitiveness lies in integrating five capabilities that Insta360 has accumulated over the years — imaging, audio, gimbals, drones and AI algorithms — into a single technical framework, rather than building a general-purpose robot from scratch.

However, for a hardware company, Cameraman will still take time, but closer at hand is data.

Qi Lu told Chuangye Zui Qianxian that Insta360 has built its own data annotation team since 2018, accumulating more than 30 types of scenes and over 10,000 hours of panoramic data over seven or eight years. The training set for just one depth prediction task alone contains 2 million images. "We realized early on that data is the core principle for building standards," he said.

Data also changed the company's attitude toward open source. Qi Lu did not shy away from internal debate: open source versus closed source is essentially about whether technology should be disclosed externally. The company ultimately leaned toward open source, on the grounds that "we hope to build a strong panoramic ecosystem through industry collaboration, and these capabilities will ultimately feed back into product competitiveness."

In January this year, Insta360 open-sourced its panoramic depth model DAP360, and the bad cases reported back by the community became fuel for improving its in-house model.Qi Lu gave an example: "If you only have one eye in front and someone comes up behind you, you cannot perform omnidirectional obstacle avoidance." In Insta360's vision, these data could also serve robot obstacle avoidance and spatial understanding, giving the camera company the potential to become a visual infrastructure provider.

The same pragmatic logic applies to large language model selection.

GO Ultra's AI voice assistant Kira uses Alibaba Qianwen in China and Gemini overseas, but local modules such as voiceprint recognition and wake words are still developed in-house. "For core technologies that are distinctive to Insta360, we prefer to develop them ourselves; for more general capabilities, we prefer to co-develop them with external vendors," Qi Lu told Chuangye Zui Qianxian.

This "in-house core plus co-developed general capabilities" model has let a hardware company maintain its product cadence amid the AI wave.

3. Leave editing to AI and weave recording into daily life

On August 7, five days before the X6's launch, Kira went live with a firmware update for the GO Ultra. Insta360 defines Kira as a "pocket AI guide" rather than an all-purpose AI assistant.

The scenarios in question include, for example, when a user's hands are occupied while cycling or surfing, asking for directions or checking the weather would actually interrupt the experience. Kira offers face-to-face translation, photo-based Q&A and voice Q&A. Complex computation is handled on the phone and in the cloud, while the device locally retains only command words, voiceprints and captured data; at 53 grams, the added power draw of the body is limited.

Insta360's logic is that hardware is never finalized in one shot: through OTA updates, older models can also unlock new features.

The bigger change is in editing. The biggest pain point of 360 cameras has never been capturing the shot, but having to edit carefully afterward. The zero-editing feature on the X6 is powered by Insta360's in-house 360 AI model, PanoMind, trained on the data mentioned above. Its AI director function can identify highlights, generate virtual camera moves, and score the editability of each clip.

When users plug in the camera to charge, it automatically selects and edits footage, with the finished video pushed to their phone the next morning. Liu Jingkang sums it up as "the user sleeps, the camera works."

The Insta360 Pro goes a step further: users set the theme and style in natural language, and AI selects the footage and structures the narrative. Insta360's cloud-based AI editing is charged at 6 yuan per clip, with an AI export rate of nearly 60%. Users pay for "automatic video creation," not for computing power or storage.

What used to be done in post-production software is being pushed into the cloud.

The road has not been easy. Qi Lu told Chuangye Zui Qianxian that photography is highly subjective, and that "a thousand people, a thousand tastes" is the hardest part of automatic editing: the model must infer aesthetic preferences from user behavior.

He also revealed that Insta360's imaging pipeline follows two paths: on-device, where computing power is limited, small models or even traditional CV handle basic needs; more demanding experiences are left to Insta360's own cloud.

At the same time, Insta360 has deliberately held back on the ultimate form of Cameraman.

"We are not going to build a truly ultimate Cameraman product," he said. In his view, GO Ultra's head-mounted accessory and X6's automatic tracking are staged validations. "A product vision cannot be achieved overnight; it must follow both technical logic and commercial logic."

Six days before the launch, Liu Jingkang discussed an even bigger shift at AGI Playground: future imaging devices will be "light at the front, heavy at the back," with the front end handling lightweight capture while the heavy lifting of editing, aesthetics and delivery is left to AI on the back end.

As stated in the 11th anniversary all-hands letter, "In the past, we made cameras see more completely; next, we will make cameras move on their own." He defined Cameraman as "an agent that takes photos by itself."

In a conversation at WAIC, he repeatedly drew a line for AI: "AI's impact on the sharing and commercial creation of images is revolutionary, but in the dimension of image documentation, AI cannot surpass it. The original impulse to document is that this moment is worth cherishing, and that part will not be replaced by AI."

This may be a camera company's deepest moat in the AI era: generative AI can synthesize any image, but it cannot synthesize “memories of things that actually happened.”

Although 43% of Insta360's internal code is now AI-generated and more than half of customer service is handled by AI, with efficiency as the theme of the organization, what it sells remains the real moments that belong to everyone.

As Qi Lu put it, the ideal camera should be imperceptible.

In ten years, Insta360 shifted the focus of competition from specs to scenarios, from scenarios to data, and now to editing. In fact, it has done only one thing: let the camera fade into everyday life and return creation to ordinary people.

       Original title : Insta360 at Ten Years of 360-Degree Imaging: Making the Camera “Imperceptible” and Returning Creation to Ordinary People