On August 22, 2023, Volcano Engine Video Cloud announced that its self-developed video codec chip has been successfully produced. It has been verified that the video compression efficiency of this chip can be increased by more than 30% compared with the industry's mainstream hardware encoders. In the future, it will serve video services such as Douyin and Xigua Video, and will be opened to enterprise customers through the Volcanic Engine Video Cloud.
Tan Dai, president of Volcano Engine, said that the amount of video data is growing explosively, and the application scenarios are also expanding. To this end, Volcano Engine and Nvidia have cooperated to open source the BMF (Babit MultiMedia Framework) multimedia processing framework, which can better support heterogeneous computing resources such as CPU, GPU, and video codec chips, and give full play to their respective advantages in different scenarios. Reduce computing, storage and transmission bandwidth costs of network video services.
Li Xipeng (left), general manager of the Asia-Pacific region of NVIDIA's development and technology department, and Wang Yue (right), technical director of the video cloud architecture of the volcano engine, jointly announced the open source BMF multimedia processing framework
High compression efficiency, high computing density
Video is a dynamic image sequence connected by pictures. Video codec is the basic technology of video application. Its function is to compress and digitally encode images, and transmit high-quality video data with the smallest possible bandwidth.
Based on the large-scale practice and polishing of video services such as Douyin , Volcano Engine integrates self-developed video codec technology into dedicated chips, and the compression efficiency is more than 30% higher than that of mainstream hardware encoders in the industry. It can be applied to video on demand and live broadcast. , image compression, XR and other business scenarios.
Compared with CPU chips with general-purpose computing capabilities, Volcano Engine's self-developed chips only support video encoding and decoding tasks, so they can have higher computing density in specific scenarios. The transcoding capability of one video codec chip server is equivalent to that of hundreds of CPU servers.
Wang Yue, technical director of the video cloud architecture of Volcano Engine, believes that “every generation of video coding is upgraded, the quality bit rate is reduced by 50%, but the large-scale video data volume increases by more than 2 times, and the iteration speed of video coding standards can no longer catch up with the video data volume. To increase the speed of growth, we must break through the bottleneck of computing power through hardware innovation.”
"With the rise of ultra-high-definition/8K VR live broadcast, real-time audio and video interaction and other increasingly rich low-latency video scenarios, it is particularly important for enterprises to ensure the quality, efficiency, and cost of video processing." Wang Yue said, Volcano Engine will continue to explore from different dimensions such as computing power, experience and interaction, and bring more cost-effective video cloud services to customers.
Video becomes the new driving force for enterprise digital growth
According to third-party data, 69% of the top 100 apps in China are equipped with video functions, 65% are equipped with on-demand functions, 51% have live broadcast functions, and 32% have real-time audio and video functions. Video applications are expanding from content and entertainment industries to pan-Internet and a large number of traditional industries.
Compared with graphic content, video applications built by enterprises have higher requirements for technology accumulation and capital costs. To this end , Volcano Engine and Nvidia have open sourced the BMF multimedia processing framework. The framework is the "base" for video deployment, which can provide enterprises with full-link scenario support from video production to consumption, and help developers build video applications at low cost.
Wang Yue introduced that the BMF framework was polished by Douyin for many years of practice, and the Volcanic Engine worked with Nvidia to deeply optimize GPU-related capabilities, enabling it to exert more powerful parallel computing capabilities.
Li Xipeng, general manager of the Asia-Pacific region of NVIDIA's development and technology department, said that NVIDIA and Volcanic Engine have always aimed to make BMF a basic platform for next-generation video processing. "In the future, we will maintain cooperation with Volcano Engine, listen to the opinions of the community and users, iterate on NVIDIA's latest software and hardware platforms, and continue to improve the BMF framework." Li Xipeng said.
Combined with the AIGC capabilities of the Volcano Engine Intelligent Creation Cloud, the Volcano Engine Video Cloud introduces AIGC capabilities such as digital humans and virtual live broadcast rooms in commodity marketing, interactive entertainment, online education, digital finance, and smart driving, helping companies reduce costs and increase efficiency, Improve user experience. Recently, the Volcanic Engine Video Cloud also used large-scale visual models to restore 4K classic Hong Kong movies for the first time, helping to further improve the image quality of old movies.
Liao Qian, head of the Volcano Engine Smart Creation Cloud, said that the Smart Creation Cloud product has provided Douyin Life Service merchants with complete AI capabilities from script creativity, video production, content release and data feedback, which has greatly reduced the video usage of enterprises and businesses. threshold. Statistics show that Smart Creation Cloud products have served more than 3,000 enterprises and merchants, synthesized more than 3.09 million videos, and the video exposure reached 1.54 billion.
Tan Dai pointed out that in the video era where the interactive experience of marketing, space, business and knowledge is changing, video cloud technology has been given an important mission because of its full-link technical capabilities from video production, video processing to video consumption. With the rise of AIGC, XR and other technologies, the integration of new technologies and video cloud will bring more innovation and value.