Alibaba new AI-based video generation tool to rival early movers

Alibaba Cloud, a subsidiary of the Chinese technology giant Alibaba (NASDAQ: BABA), has unveiled a groundbreaking AI-powered video generation tool to compete with other industry pioneers. The new tool, known as I2VGen-xl, has demonstrated exceptional capabilities in generating high-quality videos from a variety of sources. It is highlighted for its ability to create videos that are both visually striking and semantically accurate, minimizing the likelihood of errors, hallucinations, or false representations. According to statements on GitHub, the tool can produce videos based on input text, images, desired motion, subjects, and even user feedback signals. The open-source video generation codebase, called VGen, enables users to train their own text-to-video models effortlessly through a simple Python command. The repository supports compositional video synthesis, motion controllability, instruction with human feedback, and T2V scaling. It includes various pre-trained models for multiple tasks and a range of commonly used video generation tools. VGen’s advanced features are the result of extensive training on a massive dataset comprising 6 billion text-to-image pairs and 35 million text-to-video pairs, guaranteeing versatility and accuracy in various applications. In addition to releasing technical papers and an official webpage, the development team plans to introduce models specifically designed for generating videos of human bodies and an enhanced version for motion capture. Alibaba’s continuous investment in emerging technologies has propelled the company’s entry into the AI space, evidenced by the launch of Tongyi Qianwen, a large language model that competes with Meta’s Llama 2, and the introduction of “Animate Anyone,” an offering that transforms static photos into videos using Alibaba’s ReferenceNet framework. Despite facing challenges due to the ongoing semiconductor cold war between the US and China, Alibaba is determined to pioneer AI and quantum computing solutions by leveraging its advanced video generation tool.