Chinese AI startup launches new model to rival OpenAI's Sora

Shengshu AI, a Beijing-based startup, has launched Vidu Q2, its latest video-generation model designed to compete with OpenAI’s Sora.

Vidu Q2 allows users to create short AI-generated videos using up to seven reference images and supports text-to-video and image-to-video features.

The platform offers membership plans from free to a flagship tier at US$84 a month, providing faster output and more capacity.

It can generate clips up to eight seconds long in multiple aspect ratios, including synchronized dialogue and sound effects.

Shengshu said its user base reached over 30 million people in more than 200 countries, with annual recurring revenue above US$20 million.

The company recently raised hundreds of millions of yuan in series A funding led by Bohua Capital.

The launch of Vidu Q2 comes amid a global push to develop advanced AI video models, with recent releases from OpenAI, Google DeepMind, ByteDance, and Tencent.

🔗 Source: South China Morning Post

🧠 Food for thought

Implications, context, and why it matters.

Vidu Q2 reference-to-video edge hinges on unspecified API economics

The seven-reference-image system merges distinct characters, objects, and backgrounds into single clips 1. It fits identity consistency workflows when animating from stills 2.
Support covers reference-to-video, image-to-video, and text-to-video with synced dialogue plus sound effects. The News Article leaves out per-clip costs, Application Programming Interface (API) rate limits, and commercial-use terms that builders need to plan Software-as-a-Service (SaaS) integrations.
Without clear API pricing beyond consumer membership tiers at $84 per month for the flagship 1, agencies and tool makers cannot model per-output costs or margins.

Developers outside China can earn from Vidu character consistency if API access works

Independent developers and agencies outside partnerships could use global availability 1 plus reference-image control to build niche SaaS for webcomic studios, indie game cutscenes, or social media creators who need consistent characters across shots.
Clips run up to eight seconds with audio. Teams could chain sequences to form a complete arc. That enables links with scriptwriting tools or storyboard platforms that automate multi-shot stories with locked designs.
Success hinges on API access terms that the News Article does not detail. Vidu’s platform 3 might offer pay-per-use access like fal.ai. Fal.ai is a developer platform that sells hosted AI inference and media generation. Its video pricing is charged per second of output 4. Or it might require enterprise contracts that price out smaller developers.