On May 27th, CCBN (The 28th China International Radio and Television Information Network Exhibition) was grandly held in Beijing. In this "CCBN Annual Innovation Award" selection, Cloud Video Cloud relied on 160b99e4453bd9 AI editorial department . Professional solutions for the media industry won the CCBN Product Innovation Excellence Award.



With the development of technology, the innovation and application of new technologies such as 5G, 4K, and AI have become the consensus of the media industry's transformation. Facing the 986 million mobile phone netizens, it has become a development trend of the media industry to make full use of the advantages of 5G network technology and build an all-media distribution system that integrates small screens on mobile phones and large TV screens. Under the new form of communication, users are more inclined to personalized audio-visual needs, which brings new requirements and challenges to the media industry's content production timeliness and production output. How to use artificial intelligence and data technology to improve video production timeliness and production output has become one of the key tasks of the AI editorial department's media integration transformation.

Production production is the core of the entire production and broadcasting process. It connects collection, playback, and storage. Only when the production process is fully clouded can the efficiency of converged media be truly improved. In the entire converged media production and broadcast process, the newly upgraded AI editing department 2.0 combines short video production and high-bit-rate precision editing, and uses AI technology to achieve two-way interaction and integrated production of large and small screens. At the same time, the integration of cloud and end-to-end collaborative production methods accelerates the efficiency of the entire content production.

From the practical operation level, the AI editorial department is realized through three subdivisions: true 4K cloud production, 4K video enhancement, and cloud media short video production.

First of all, in 4K cloud production infrastructure, the main innovation made by the AI editorial department is the introduction of SCC high-performance computing and CPFS parallel file system. In response to the high-performance requirements of cloud production, low-latency interconnection between computing nodes and storage nodes is realized to meet a better production experience, while avoiding virtualization losses and further improving performance. In response to the high storage capacity and bandwidth requirements of cloud production, CPFS provides linear expansion throughput capacity with storage nodes, with a maximum throughput capacity of 300GB/s in a single cluster. The AI editorial department provides a wealth of resources adapted to the media business, including various personalized hardware devices such as FPGAs and GPUs, to meet different business demands. In addition, the AI editorial department has made a lot of optimizations for the GPU, including GPU virtualization, sharing, and redization, which can greatly increase the utilization rate of the GPU and reduce the cost of cloud production.

Secondly, in 4K video enhancement , the AI editorial department mainly uses three core technologies: video super-division, video frame insertion, and color enhancement to comprehensively improve the picture quality of 1080p and lower quality videos. All three aspects of color meet the requirements of ultra-high-definition video. The 4K enhanced capabilities provided by the AI editorial department not only meet the professional standards, that is, the standard definition in the "4K Ultra HD TV Program Production Technology Implementation Guide" by the General Administration, but also meet the professional technical review and monitoring requirements in the industry. At the same time as the parameter standard, the subjective visual experience has been significantly improved, including detail restoration, noise suppression, smooth picture, video brightness, color contrast, and preservation of light and dark details. In terms of AI video super-segmentation, the AI editorial department uses deep learning to perform lossless enlargement of the picture, which can achieve a 4x increase in resolution, restore and restore details such as portraits and scenes, while suppressing block noise, compression noise, etc. The picture quality is damaged, making the picture sharper. The AI editorial department has innovatively proposed the enhancement of key parts to achieve the most effective results at the least cost, such as focusing on optimizing the portrait face, and achieving better restoration and restoration of human skin texture and hair details.

Finally, in the cloud-based short video production , combined with the current short video hotspots and user preferences, starting from the short video production standpoint, the AI editorial department has the following 5 core advantages :

  1. The AI editorial department provides powerful prediction and evaluation capabilities to help editors decide what kind of short videos to make and make targeted productions.
  2. The AI editorial department provides powerful multi-modal labeling capabilities to redefine the content of the video. It comprehensively utilizes multiple AI capabilities such as images, voices, and text to systematically generate labels.
  3. Aiming at the pain points of retrieval, the AI editorial department provides powerful multi-modal retrieval, which includes text search, image search, video search, and support for natural speech understanding, knowledge graphs and other full-process multi-modal retrieval capabilities.
  4. Provide massive, multi-category video templates to produce a large number of professional short videos in a short time.
  5. The AI editorial department provides a wealth of professional intelligent auxiliary tools, which can meet the needs of fast short video production scene-based.



    Since the release of the AI editorial department 2.0, the People’s Daily Technology Department has combined with Alibaba Cloud Video Cloud to use this capability to with the content of the two sessions, using Alibaba Cloud’s AI technology, through the massive data training of the People’s Daily, and the business of the People’s Daily. AI application innovation in the scene. The AI editorial department "will watch, listen, understand, and check", which can realize the automation, batch and intelligent production of various new media products, and realize functions such as multi-modal retrieval, intelligent clearing, and intelligent video production. Greatly improve the efficiency of news production, and comprehensively assist the People's Daily to report on the two sessions. After continuous iteration and improvement, on the eve of the and CPPCC in 2021, Alibaba Cloud Video Cloud will comprehensively upgrade the functions of the 160b99e4453e79 AI intelligent editorial department People’s Daily, integrate five new AI capabilities, and help newspapers and media to report on the NPC and CPPCC. Improve quality and efficiency.

    In the past , the AI editorial department Radio and Television used professional equipment locally to produce high-definition and standard-definition video.

    is now . The AI editorial department uses IP, intensification and virtualization to decouple professional equipment and production software.

    future , through the AI editorial department, Alibaba Cloud Video Cloud will collaborate with more partners to use big data and artificial intelligence technology to provide remote, interactive, and intelligent cloud production capabilities to accomplish more imaginable Business scenarios help the digital transformation and upgrading of the media industry.
"Video Cloud Technology" Your most noteworthy audio and video technology public account, pushes practical technical articles from the front line of Alibaba Cloud every week, and exchanges and exchanges with first-class engineers in the audio and video field. The official account backstage reply [Technology] You can join the Alibaba Cloud Video Cloud Technology Exchange Group, discuss audio and video technologies with the author, and get more industry latest information.

CloudImagine
222 声望1.5k 粉丝