As a provider of intelligent video and audio solutions, Cam SDK is committed to providing the most cutting-edge overall video and audio solutions for industry institutions such as mobile Internet, smartphones, smart hardware, traditional radio&television and communication operators. At the same time, it provides developers and industry partners with efficient and stable underlying technology support to help partners achieve business innovation and create user value.
On October 31, 2020, LiveVideoStackCon 2020 Audio and Video Technology Conference was held in Beijing Liting Huayuan Hotel. The conference focused on audio and video technology and globalization, AI and multimedia processing, education and multimedia, video codec, server development, optimization, and other related audio and video technology and industrial applications. It brought together more than 500 senior technical experts and top scholars from pan-entertainment, education, video conferencing, social networking, games, e-commerce, finance, IoT, and other industries.
Mr. Hou Kang, the leader of the AI research group, was invited to attend and share the keynote speech “Technical Framework and Challenges of AI Synthetic Virtual Characters.” The speech started from the AI synthetic virtual character technology framework and introduced the AI synthetic virtual character technology background. This technology can generate a video of an anchor by inputting a piece of text. Besides, it further introduced the differences and advantages & disadvantages between different technical paths of AI synthetic virtual characters, the design and optimization ideas of facial expression and mouth prediction model, 3D face reconstruction and rendering technology, confrontation generation network, video frame rendering and fusion and other key technologies. At the same time, solutions were also shared for such difficulties as mouth shake and distortion, facial detail distortion in GAN rendering results, and smooth transition between frames before and after the final video.
The AI synthetic virtual anchor of Cam SDK is a combination of artificial intelligence and computer graphics. After the computer gets the speech or text content, the virtual image’s face is driven, rendered, and integrated through the computer graphics synthesis technology. After the massive face data is generated and the anti-network training is carried out, the realistic expression is restored, and the virtual anchor is the same as the real person. The technology is widely used in news broadcast, virtual customer service, Internet teaching, and other fields.
At present, the product matrix of Cam SDK includes SDK of HD video shooting and in-depth editing, intelligent video production platform, intelligent editing solution, smartphone solution, face + beauty/stickers solution, cloud editing, virtual anchor, customized design of video packaging, etc. The products cover the mobile end, PC end, and Web end. The whole process function from shooting to editing realizes the current gameplay of all mainstream videos. Flexible technology architecture, multi-threading and GPU, CPU collaborative preprocessing mechanism, natural language processing algorithm, and other technologies have been unanimously recognized by manufacturers in various industries. It has carried out technical cooperation with Xiaomi, OPPO, ViVO, ASUS, Cheetah Mobile, Bilibili, SINA, KEEP, and other top customers in various fields. It also cooperates with traditional enterprises such as State Grid and Broadcasting to promote Toutiao APP and media convergence in government administration and enterprises. Cam SDK is also expanding its overseas market, filling the gap of intelligent video and audio solutions in the overseas market. So far, it has provided in-depth technical services with local Internet companies in the United States, South Korea, India, Singapore, and other countries, radiated tens of millions of overseas users, and helped partners realize business innovation and create user value.