Alibaba has now made Wan 2.1, an image-generating artificial intelligence model, open source. To improve the accuracy of its image and video generation, the Chinese e-commerce giant has unveiled four versions of Wan 2.1: T2V-1.3B, T2V-14B, I2V-14B-720P, and I2V-14B-480P.
The “14B” designation signifies that these models utilize 14 billion parameters, enabling them to handle significantly more data and produce more refined results. It offers capabilities in text-to-video, image-to-video, video editing, text-to-image, and video-to-audio.
The I2V-14B and T2V-14B models from Alibaba both produce videos at 480P and 720P, with the T2V-14B standing out as the only model capable of generating videos containing both Chinese and English text.
In contrast, the T2V-1.3B is designed for consumer hardware, requiring only 8.19 GB of VRAM on an RTX 4090 to generate a five-second 480P video in four minutes.
Publicly releasing the AI model is expected to drive widespread use and intensify the competitive landscape of artificial intelligence. This announcement by Alibaba is in line with a previous action taken by Chinese company DeepSeek.
Wan 2.1 is meant to be a direct competitor to OpenAI’s Sora. Alibaba initially unveiled its advanced video and image generation AI in January, later rebranding it from Wanx to Wan.
Additionally, on Tuesday, Alibaba previewed QwQ-Max, a reasoning model, with plans to open-source the full version upon its official launch. It also announced plans this week to invest at least 380 billion yuan ($52 billion) over the next three years to bolster cloud computing and AI infrastructure.
