Tencent Cloud Video Services: Technical Insights on Live Streaming, Codecs, and AI Enhancements

During the pandemic, remote work and online education became essential, leading to a surge in video‑conference and live‑streaming traffic. LiveVideoStack interviewed Li Yutao, General Manager of Tencent Cloud Video Communication, to discuss how the platform handles high concurrency, high availability, and high performance challenges.

Li explained that Tencent Cloud Video has expanded its product matrix with lower‑latency fast‑live, slow‑live for large‑scale events, cloud editing (Tencent Cloud Clip) and cloud switcher, as well as a family of AI‑enhanced video products. The video cloud now serves as a central platform for WeChat, WeChat Work, Tencent Meeting, and Tencent Education.

To support overseas users, Tencent has deployed multi‑regional data centers (Hong Kong, Thailand, Singapore, Germany, Toronto, Silicon Valley, Russia, South Africa, Korea, etc.) and built over 1,300 CDN nodes in more than 50 countries, providing over 100 Tbps of bandwidth and 200+ acceleration points.

For real‑time communication, Tencent Meeting relies on a cloud flow‑control engine that combines classic signal processing, psycho‑acoustic models, and deep‑learning techniques. It uses TRTC + WebRTC fast‑live, RTMP, and CDN fusion to dynamically switch streams based on concurrency thresholds, achieving sub‑300 ms latency for TRTC and 500 ms–5 s for fallback paths.

Regarding transport protocols, Tencent supports SRT, which offers live‑scene‑optimized bitrate pacing and configurable packet loss handling, outperforming QUIC, RTMP, and HLS in long‑distance, high‑latency links.

On the codec front, Tencent joined the AOMedia Alliance to promote AV1, and its live‑streaming and VOD services already support AV1/AVS2. Development of VVC/AVS3 is underway and expected to be released in 2020.

AI‑driven technologies include the “Smart Dynamic Encoding” system that performs scene‑aware encoding, deep‑learning‑based quality enhancement, and bandwidth‑saving up to 30‑40 % compared with open‑source solutions. The “Intelligent Content Understanding” suite (video analysis, automatic editing, moderation) powers applications such as automatic clipping of CCTV’s 70‑year anniversary parade, real‑time highlight generation for esports, and content‑safety checks.

Looking ahead, Li predicts that short‑term demand will continue to shift offline content to online platforms (remote work, education, virtual commerce). Long‑term growth will be driven by online collaboration, AI‑generated content, virtual anchors, and immersive experiences like VR‑based commerce.