China’s text-to-video AI race has a new entry point: Vidu, a model introduced by Shengshu Technology and Tsinghua University. The launch places another serious contender into a field shaped by OpenAI’s Sora, while also showing how far competitors still have to go.
According to the source article, Vidu can create a 16-second HD video at 1080p resolution with a single click. Shengshu Technology says the model is "very close" to the level of OpenAI’s Sora, though the article also notes that this claim comes with important limits.
What Vidu can do
Vidu was unveiled at the Zhongguancun Forum 2024 in Beijing by Chinese AI company Shengshu Technology and Tsinghua University. Its core pitch is straightforward: turn text into short HD video quickly, with a level of scene continuity that its developers describe as "exceptional consistency."
In plain terms, that consistency means the frames in a video should connect in a way that appears logical rather than random. For text-to-video AI, this matters because a clip is not just a collection of attractive still images. It needs characters, objects, movement, and visual logic to hold together over time.
The 16-second duration is the clearest published capability in the source. The 1080p resolution also gives Vidu a practical benchmark, because video generation is judged not only by whether an idea appears on screen, but also by how usable the resulting clip looks.
How it compares with Sora
The source article frames Vidu directly against OpenAI’s Sora. That comparison is unavoidable because Sora has become a reference point for high-end text-to-video generation, and Shengshu Technology itself says Vidu is "very close" to that level.
Still, the most concrete difference is length. Sora can generate continuous videos of up to one minute, while Vidu currently manages 16 seconds. That gap matters because longer clips make it harder for a model to maintain a coherent subject, setting, and sequence of events.
The article also says Vidu appears to lag Sora in quality despite the confidence of its developers. The conclusion is not that Vidu is insignificant, but that its current public description does not yet put it on equal footing with OpenAI’s video debut.
- Vidu: 16-second HD video at 1080p resolution with a single click.
- Sora: continuous videos of up to one minute.
- Key gap: Vidu’s shorter output makes the comparison less favorable for now.
The panda and dragon claim
One of Shengshu Technology’s more specific claims is that Vidu can better "understand and generate Chinese elements such as the panda and dragon" compared with Sora. The source article treats that claim cautiously, saying it has yet to be proven in practice.
That point is important because cultural specificity is not just a decorative feature in generative AI. If a model is presented as stronger at producing certain visual elements, the evidence has to show that it can represent them reliably and consistently in actual generated videos.
The source also says Shengshu Technology reports that the model’s core architecture was developed in September 2022, before the launch of Sora. That detail supports the company’s position that Vidu is not simply a reaction to Sora, while the article still judges the model by visible performance.
Why the launch matters
Vidu’s arrival shows China’s ambition to compete with leading US companies such as OpenAI in generative AI models. The source article is clear that catching up, or moving ahead, will require a significant increase in performance.
One possible factor mentioned is limited access to GPUs in China compared to OpenAI. The article presents this as a reason Vidu may be behind Sora, not as a full explanation. The broader takeaway is that text-to-video AI depends not only on model design, but also on the resources needed to train and scale such systems.
Sora is expected to be released this year, and OpenAI plans to scale the model further. Details on pricing and generation times are not yet known. That leaves the competitive picture unfinished: Vidu has been shown as a serious Chinese effort, while Sora’s public rollout and future scaling remain key unknowns.
For now, Vidu is best understood as a meaningful step rather than a finished answer to Sora. It brings 16-second 1080p text-to-video generation into the spotlight, claims strength with Chinese elements, and signals that Shengshu Technology and Tsinghua University want a place in the top tier of generative AI video. But based on the source article, Sora still holds the clearer advantage in video duration and apparent capability.