Tencent launches the open-source HunyuanVideo-I2V, transforming images into high-res videos using AI. The tool is widely available on platforms like GitHub, enriching video content with sound effects. This innovation fuels competition in China’s AI video-generation market, alongside contenders like Kuaishou and Alibaba, following OpenAI’s Sora model.
Tencent Holdings has launched its new open-source model, HunyuanVideo-I2V, invigorating the competition in AI video-generation technology in China. This innovative image-to-video model allows users to convert a static image into a high-resolution video clip, reaching 720 pixels and featuring up to 129 frames in just five seconds via short text prompts.
Available through platforms like GitHub and HuggingFace, this tool signifies Tencent’s commitment to advancing AI. Users can enrich their videos by adding lip-synched voice and background sound effects, vividly illustrated by a demonstration of Albert Einstein munching an apple with a matching crunch sound effect.
As the market heats up, Tencent enters a landscape populated by formidable competitors such as Kuaishou Technology, TikTok’s ByteDance, and Alibaba’s cloud services. Since OpenAI’s groundbreaking Sora model was introduced over a year ago, the demand for home-grown video-generation products in China has surged.
Previously, Tencent unveiled the HunyuanVideo foundation model in December, promoting it as the largest open-source model for video generation worldwide, boasting over 13 billion training parameters. This model’s accessibility for free to businesses and individuals marks a significant move in democratizing AI video creation in China.
Tencent’s launch of the HunyuanVideo-I2V model showcases its ambition to lead the AI video-generation market in China. As competition intensifies with players like Kuaishou and Alibaba, developers now have powerful, accessible tools at their disposal. The democratization of AI creates exciting opportunities for creativity and innovation in digital content creation.
Original Source: www.scmp.com
