Chinese AI startup MiniMax has unveiled Video-01, its first AI model capable of generating high-resolution videos from text prompts.
CEO Yan Junjie presented Video-01 at a company event, sharing some key features. The model supports a native resolution of 1,280 x 720 pixels at 25 frames per second and allows for virtual camera control.
Currently, videos are limited to six seconds, but MiniMax plans to extend this to ten seconds in the near future. The company hasn't released technical details about the model's architecture or training data.
MiniMax intends to release regular updates for the video generator. Plans include allowing image input alongside text and giving users more control over the final output.
Users can try Video-01 for free at hailuoai.com/video after registering with a mobile number, which also works for non-Chinese numbers. MiniMax also provides an API for developers. Yan emphasized that MiniMax is currently focused on making the technology widely available rather than commercializing it.
Video quality and capabilities
Videos generated with Video-01 are still easily identifiable as AI-made. They appear slightly oversaturated with smooth surfaces, resembling heavily filtered footage. Video-01 offers various styles, including anime, CGI, and video game graphics. The model shows relatively few image errors or artifacts and even seems capable of displaying text in videos.
Video: Video-01 prompted by MiniMax
Video: Video-01 prompted by MiniMax
Video: Video-01 prompted by MiniMax
Video: Video-01 prompted by MiniMax
After registering, MiniMax asks users to accept its terms of use, which include not generating illegal content. The company warns against using the tool to spread rumors, violate privacy, or share illegal information.
However, there seem to be few limits to the AI's capabilities, as it can generate videos of famous personalities and even political figures such as Donald Trump and Vladimir Putin. The tool blocks pornographic or explicit scenes and adds a discreet watermark to the bottom right corner of the generated videos.
Video: Video-01 prompted by THE DECODER
Company background and funding
MiniMax was founded in late 2021 and offers a large language model and a text-to-speech model, in addition to the new video generator. A recent funding round led by Alibaba raised about $600 million, including an investment from Tencent, valuing MiniMax at at least $2.5 billion.
While OpenAI's Sora is still considered the industry benchmark for video AI, at least according to demos and reports, numerous companies, especially from China, have also entered the video AI market. Unlike OpenAI, these companies, including KLING, Vidu, and Jimeng AI, have made their tools available to users.
