Recently, Modelers officially launched two open source multimodal models, Step-Video and Step-Audio developed by Step Yuexingchen, marking a major breakthrough in AI technology in the fields of video generation and voice interaction. These two models not only provide developers with powerful tools, but also bring more innovative possibilities to enterprise users.
Step-Video, full name Step-Video-T2V, is currently the world's largest open source video generation model with a parameter volume of up to 30 billion. It can directly generate high-quality videos with 204 frames and 540P resolution, and surpass the existing top open source video models on the market in terms of command compliance, motion smoothness, physical rationality and aesthetics. The launch of this technology will undoubtedly bring revolutionary changes to the field of video creation.
Step-Audio is the industry's first voice model that can generate a variety of emotions, dialects, languages, singing styles and personalized styles. The release of this technology not only enriches the diversity of voice interactions, but also opens up new directions for AI voice applications. Whether it is virtual assistant, smart customer service, or voice synthesis creation, Step-Audio will bring users a more natural and personalized experience.
It is worth mentioning that these models are adapted based on Huawei Ascend CANN heterogeneous computing architecture and Ascend server, ensuring its powerful computing power and efficient operational performance. Developers and enterprise users can easily download and experience these models through the Mole Community. In order to further lower the threshold for use, the Mole community also provides free computing power support, so that users can perform model inference online without the need for complex environment construction, and quickly verify their AI solutions.
Jieyuexing's open source model has attracted the attention of many industry benchmark companies, including Tianshu Smart Chip, Alibaba Cloud, Volcano Engine, TCL and other manufacturers from all walks of life have connected to this open source ecosystem. This not only reflects the technological leadership of these models, but also provides a new platform for technological cooperation and innovation in the industry.
In the future, Jieyuexingchen plans to launch a new Tusheng video model in March to further enrich its product line. This move will further enhance its competitiveness in the field of multimodal AI and bring more innovative solutions to users.
The cooperation between Huawei Ascend and Jie Yuexing not only expands the application scenarios of multimodal AI models, but also provides developers with more powerful tools and promotes technological progress in the entire industry. With the continuous development of AI technology, we look forward to seeing more similar technological breakthroughs and bringing more innovation and changes to all walks of life.