Realtime AI News
CINMOORE's DB2 doorbell runs on Alibaba's Tongyi Qianwen to explain, not just record
CINMOORE, the smart camera brand under Hangzhou-based Yanji Microelectronics, says its new DB2 4G dual-camera doorbell now runs on Alibaba's Tongyi Qianwen large model. The model powers AI event summaries, visitor role recognition and a daily highlight reel, shifting the device from recording what happens at the door to explaining it.
CINMOORE has announced that its new-generation 4G dual-camera radar-vision fusion smart doorbell, the DB2, is now officially connected to Alibaba's Tongyi Qianwen large model. According to the company, three core capabilities — AI event summaries, visitor identity recognition and a daily highlight summary — are all provided by Qianwen. The doorbell's job has shifted from recording what happens at the door to telling users what happened and who showed up.
The change targets a gap that a decade of hardware progress did not close. Over that period, mainstream doorbells moved from million-pixel to 5-megapixel sensors, from single to dual cameras, from black-and-white to full-colour low-light night vision, and from Wi-Fi to 4G connectivity.
Yet the everyday problem persists. A typical home entrance can trigger dozens of events in a single day — deliveries, food couriers, neighbours passing by, a cat in the corridor. Users receive dozens of notifications and still have to dig through dozens of clips to answer simple questions: who was that, did my parcel arrive, did someone linger by the door.
A CINMOORE representative framed the shift bluntly: the industry keeps competing over who films more clearly, but what users actually need is not a clearer picture so much as a clearer answer. Seeing, the company says, is the camera's job; understanding is what counts as intelligence.
Technically, the DB2 splits the work. On the sensing side it uses CINMOORE's own radar-vision fusion approach: a 24GHz millimetre-wave radar first detects movement at the door, the camera then re-checks the frame, and recording starts only when the image is meaningful. Compared with conventional PIR infrared sensing, the company says this reduces false alerts caused by sunlight, heating and pets while lowering power draw.
The cognitive layer is handled by Qianwen. Radar and camera establish that something happened at the door; the model works out what it was, who was involved, and how to say it in a single sentence.
That translates into three features. Visitor identity recognition moves beyond a generic "person detected" alert to distinguish the apparent role or appearance of a caller, such as a courier or food delivery worker. AI event messages generate a one-line summary under each event — for example, a delivery worker leaving a parcel at 18:42 — so users can grasp the situation without opening the clip. The daily highlight summary stitches a day of door activity into a single video, a feature the company says has been well received.
On connectivity, the DB2 carries a built-in 4G module so it can run independently of home Wi-Fi, aimed at metal security doors, older buildings and homes where the router sits far from the entrance. Its two cameras are stacked vertically, the upper one watching visitors and the lower one watching packages left on the ground.
For solo residents, families with children and households with older adults, the point is less about recording than removing uncertainty. CINMOORE says the DB2 will go on sale on Douyin on September 19, with wider online channels to follow.
Why it matters
Embedding Qianwen directly into a consumer security device shows Chinese model vendors competing on concrete deployments rather than benchmarks. For the doorbell category, it also shifts the differentiator from sensor specs toward how much thinking the device does on the user's behalf.
Nearby Updates
All09/15, 10:58
Miyang Technology Launches First 'Relational Productivity Agent' Product at Shanghai Bund Summit 2026
At the 2026 Shanghai Bund Summit, Miyang Technology unveiled what it calls the first shipping product in a new relational productivity agent category, presenting the launch as a new model for human-agent collaboration. The public information so far covers positioning only, with no feature list, customers or availability date disclosed.
09/15, 10:17
Seven PhD students trained a 7B LLM from scratch in three months and open-sourced everything
A team of seven PhD students trained a 7B-parameter language model from scratch in three months and open-sourced the code, training data, and training logs. Hundreds of agents took part in the research workflow, feeding the “AI building AI” discussion.
09/15, 11:45
ByteDance CEO Liang Rubo says the company will invest more in the enterprise market after integrating Doubao, Feishu, and Volcano Engine
ByteDance CEO Liang Rubo said the company will increase investment in the enterprise, or To B, market after integrating Doubao, Feishu, and Volcano Engine. The statement places a foundation model, a collaboration suite, and cloud infrastructure on the same enterprise-facing line.
09/15, 12:00
Anthropic says it blocked AI misuse that could have supported biological weapons
Anthropic says it blocked misuse of its AI that could have supported biological weapons, according to a report carried by The Daily News of Jacksonville, NC. The disclosure highlights the dual-use risks of frontier models and puts corporate abuse-monitoring and interception practices in the public spotlight.