“When those subtitles started popping up line by line, I felt the wall standing between me and the world just crumbled.” A hearing-impaired user wrote in a social media post, “I had gotten used to ‘missing out’—missing flash sales in the live stream, missing the host's product explanations, missing so much of the vibrant sound of this world.” That was until she stumbled into a live stream on Kuaishou-W where real-time subtitles appeared.

This post brought Kuaishou's 'Low-Volume Real-Time Subtitles' feature for live streams into the spotlight for many, highlighting a previously overlooked reality: approximately 20% of users on the platform habitually watch live streams on mute or at very low volume. For them, subtitles are not merely a nice-to-have addition; they are the essential gateway to truly "hearing" the live stream world.

Kuaishou's e-commerce division developed this feature, which leverages AI speech recognition technology. When a user's phone is on mute or low volume, the system automatically transcribes the host's commentary into on-screen subtitles in real time. Users can also manually toggle the feature on or off, freely drag the subtitle position, and adjust the font size to minimize disruption to their viewing experience. Research indicates that low-volume subtitles have evolved beyond an "accessibility feature"; they tangibly enhance user engagement, comprehension, and shopping decision-making efficiency within the live stream environment.

Volume Can Be Low, But Information Shouldn't Be Absent

"Not wanting to disturb others" is one of the many scenarios where users find the low-volume subtitle feature invaluable. A recent survey of one hundred users revealed that 33.0% use it when family members are resting or children are doing homework or sleeping; 23.5% use it in quiet settings like the office or a library; and 31.3% stated they "use it whenever they watch a live stream." One user mentioned, "Sometimes when I'm putting my child to sleep, I turn the sound off. The subtitles are great—I can know what the host is saying without any audio." In these responses, the value of low-volume subtitles is concrete: they allow users to preserve a personal viewing moment without inconveniencing others. The environment can be quiet, yet users can still follow the host's讲解—this is how low-volume subtitles help users stay "immersed."

On a deeper level, the feature establishes a new information channel for hearing-impaired users, those with auditory difficulties, and some older users. A hearing-impaired user provided feedback: "Because I am hearing impaired, I really need to read the text. This feature is very convenient; I can chat with the host now, no need to listen." Another user said, "When there were no subtitles, I felt anxious and frustrated, not understanding what was being said. Now with real-time subtitles, my stress is reduced." A line of subtitles solves more than just the problem of "sound being too soft"—it enables users who previously struggled to engage with the live stream context to see what the host is saying, understand the product being discussed, and even participate in interactions and purchases more naturally. This is how low-volume subtitles help users "hear."

There is another easily overlooked value: helping users "understand." Hosts in live streams come from different regions with varying accents and speaking paces. A user from Tangshan noted in the survey: "Some southern accents are a bit hard for me to understand. With these subtitles, I realized what they were saying was actually quite good." The subtitles not only compensate for the lack of audio volume but also reduce information errors caused by accents, background noise, or the listener's auditory state. Some users have even discovered treasure hosts they would have otherwise scrolled past and missed.

Furthermore, the subtitle feature effectively enhances the live stream viewing experience across multiple scenarios, boosting immersion and practicality. For instance, one user shared, "Sometimes when accompanying my child to sleep, I turn off the audio. Relying on subtitles allows me to clearly follow the host's讲解." Another user mentioned, "After my child falls asleep at night, I want to watch a live stream nearby and maybe place an order. Real-time subtitles let me do this without turning on the sound, so I don't disturb my child's rest and don't miss key information."

From enabling users to "hear" to helping them "understand" and facilitating an "immersive" viewing experience, the low-volume subtitle feature is promoting more equitable information flow within live streams. User research shows that over 80% of users are satisfied with this functionality.

From Host's Speech to User's Screen, Achieving Instant Text Display

Real-time subtitles might seem like just a line of text at the bottom of the screen, but behind them lies a comprehensive test of real-time processing, recognition accuracy, and large-scale concurrency capabilities. The acoustic environment of e-commerce live streams is extremely complex—hosts shouting, assistants interjecting, background music, and promotional sound effects all blaring simultaneously, with hosts frequently moving around, changing distance, causing constant audio variation. Kuaishou e-commerce's real-time ASR (Automatic Speech Recognition) capability, utilizing dynamic VAD (Voice Activity Detection), hotword mining, fine-tuning on millions of hours of e-commerce live stream data, and complex scene voice separation, continuously reduces the Character Error Rate (CER). Its recognition accuracy in e-commerce live stream scenarios is significantly ahead of industry standards.

In terms of response speed, Kuaishou e-commerce has built a dedicated real-time subtitle service pipeline that supports tens of millions of concurrent viewers. The entire end-to-end process, from the audio stream to subtitle display on the user's device, takes only 400-500 milliseconds. Technical solutions ensure there is no perceptible delay on the user's end, truly achieving "speech-to-text instantaneously." This isn't an abstract concept of "real-time"—it's when the host says "three, two, one, link is live," and the subtitles have already "popped" up at the very first moment.

Enhanced Experience Ultimately Drives Better Business

For e-commerce live streams, whether users can fully understand product information directly impacts their purchasing decisions.

The low-volume subtitle feature reduces information loss when watching on mute, helping users more accurately capture key details such as price, specifications, origin, and discount timings.

Data from Kuaishou e-commerce shows that the low-volume real-time subtitle feature has been fully launched and has driven a significant increase in Gross Merchandise Volume (GMV) for e-commerce hosts' content sessions. Building on this, the iteration of a new "Keyword Highlighting" feature has further boosted GMV growth. Over the past half-month, the subtitle feature has achieved approximately 72 million daily subtitle exposures, benefiting over 8 million users. As its coverage expands, subtitles are becoming a more universally accessible foundational capability within the live stream experience.

From Kuaishou e-commerce's perspective, the value of technology lies not only in efficiency gains but also in making more genuine needs visible, understood, and served. In the future, Kuaishou e-commerce will continue to optimize product capabilities based on real user scenarios, making ongoing iterations in areas like real-time subtitles, keyword comprehension, and subtitle interaction, ensuring every user can participate in live stream shopping more comfortably.

Starting with a single line of subtitles, the flow of information in live streams is becoming more equitable. The volume can be low, but with AI, the content is never quiet—AI ensures good content is seen!