Feishu has partnered with Anker Improvements to collectively launch an clever audio-recording machine known as the “AI Recording Bean,” marking the primary {hardware} collaboration in Feishu’s historical past.
The machine is a bean-shaped, round recorder weighing simply 10 grams, outfitted with a twin MEMS microphone array. When used with its charging case, the whole weight involves about 48 grams. It helps each Bluetooth and Wi-Fi transmission, and is designed for unobtrusive, always-on recording. 
Since abroad startup Plaud gained traction in 2025 with its AI recording card, a number of Chinese language corporations have rolled out comparable merchandise. Nevertheless, it stays uncommon for a significant platform firm to straight enter the AI recording {hardware} house.
Based in 2017, Feishu has by no means launched a {hardware} product earlier than. A number of sources near the corporate revealed that the AI Recording Bean is taken into account an exploratory mission internally, with Feishu’s product staff accountable primarily for the software program growth.
Since 2024, a wave of recent AI {hardware} type components has emerged, together with recording playing cards, recording beans, AI pocket cameras, and AI glasses. All intention to determine AI interplay entry factors past smartphones, embedding AI extra deeply into on a regular basis consumer situations.
Reasonably than following the market’s dominant card-style design, Feishu and Anker opted for a smaller bean-shaped type issue as a deliberate differentiation technique. Card-style gadgets profit from magnetic attachment to smartphones however are bulkier and extra restricted in sporting situations. The bean-shaped design, against this, resembles latest AI companion gadgets, emphasizing light-weight, all-day wearability.
Customers can clip the machine to a collar or pocket edge, or connect it to a telephone utilizing magnetic equipment. This design lowers the psychological and bodily friction of recording, permitting customers to begin recording effortlessly throughout conferences, consumer visits, and even each day commutes—with out pulling out a telephone or opening an app.
One other notable characteristic is real-time summarization. Regardless of its compact dimension, the AI Recording Bean permits customers to view dwell captions and AI-generated summaries throughout recording, with help for a number of visible codecs corresponding to bar charts. This permits sooner and extra versatile entry to contextual info.
In line with Clever Emergence, when paired with its charging case, the machine delivers over 32 hours of complete battery life, provides 8GB of storage able to holding roughly 250 hours of audio, and helps quick charging—10 minutes of charging permits as much as two hours of recording. These specs place it firmly inside mainstream efficiency ranges for comparable merchandise.
The product additionally displays a broader {industry} pattern: the race to regulate nearer contextual entry factors. Since 2025, advances in basis fashions—significantly in multimodal capabilities—have sparked renewed momentum in voice-based AI merchandise.
In China, established gamers corresponding to iFlytek and Sogou have upgraded AI voice enter strategies, whereas ByteDance’s Doubao has launched its personal AI-powered enter methodology. In the meantime, AI apps like Remio and Medot give attention to quietly managing contextual info throughout work and life situations.
Whether or not by way of AI voice enter or AI {hardware}, these efforts converge on a shared purpose: turning into the reminiscence gateway for customers’ AI assistants. The extra contextual knowledge a product captures, the extra customized its AI companies can change into—and the stronger its aggressive moat.
The AI Recording Bean is designed to cowl your complete workflow from throughout a gathering to post-meeting follow-up. Customers can view AI-generated key factors in actual time, and after the assembly, the system produces structured summaries together with speaker roles, motion gadgets, and key choices.
All recordings are robotically saved within the Feishu data base, the place AI converts audio into textual content and additional into structured info. Customers can then work together with historic content material conversationally—looking, querying, and even producing new content material.
For instance, a consumer can ask, “What considerations did the consumer increase in final week’s assembly?” and the system will extract and reply primarily based on previous recordings. This end-to-end protection considerably reduces post-processing workload.
Feishu’s product philosophy is deeply formed by ByteDance’s excessive knowledge-density organizational tradition. Since its launch in 2017, Feishu has pursued a differentiated technique by first serving high-tech and web corporations, excelling in paperwork, conferences, and transcription to construct early credibility.
That technique has translated nicely into the AI period. Feishu enjoys robust adoption amongst AI-native corporations, together with DeepSeek, MiniMax, Moonshot AI, and Zhipu AI, a lot of which adopted Feishu as their collaboration platform at an early stage.
These knowledge-intensive industries—characterised by frequent conferences and heavy documentation—overlap intently with the goal customers of AI recording {hardware}. From this attitude, Feishu’s entry into AI {hardware} seems each logical and well timed, enabling deeper integration with core merchandise corresponding to Feishu Docs and Feishu Bitable.
Person habits and ache factors from these clients straight formed the product logic of the Recording Bean: distinguishing audio system in assembly minutes, robotically extracting motion gadgets, and enabling searchable historic information—all among the many commonest calls for Feishu encounters.
A prevailing view within the AI recording machine market is that whereas {hardware} designs may be shortly replicated, product expertise gaps are pushed by long-term engineering refinement. This contains steady optimization of assembly templates, post-processing workflows, and quite a few operational particulars that require sustained accumulation of experience.
Right here, Feishu holds a transparent benefit. The AI Recording Bean is powered by ByteDance’s proprietary Doubao massive mannequin, which is taken into account industry-leading in multimodal understanding. Accuracy in speech recognition and comprehension of advanced conversations straight impacts transcription high quality and clever summaries—areas the place startups might battle to catch up within the brief time period.
Supply: 36Kr
Elevate your perspective with NextTech Information, the place innovation meets perception.
Uncover the newest breakthroughs, get unique updates, and join with a world community of future-focused thinkers.
Unlock tomorrow’s traits right now: learn extra, subscribe to our publication, and change into a part of the NextTech neighborhood at NextTech-news.com
