China's AI video race just got louder. Shengshu Technology — the Tsinghua University spinout behind the Vidu video platform — has launched Vidu Q4 Preview, the first public preview of its next-generation flagship AI video model. The headline promise: performances that actually feel acted.
What Q4 Preview brings
According to Shengshu's announcement, Q4 Preview is built for professionals who produce video at scale — short drama companies, advertising agencies, social content creators and film production teams. The upgrades focus on expressiveness: reference-voice support keeps a character's voice consistent across scenes, with more natural emotional delivery and genuinely lifelike acting. On the technical side, the model outputs cinematic video in 2K or 4K with 10-bit color, so footage holds up in professional workflows.
Control is the other big theme. The model supports up to 15 image references and up to three audio references — a major step for creators who need the same character, props and settings to stay consistent across sequences. It's available now through Vidu's web product and API platform, with launch pricing starting at $0.014 per second.
Why it matters
AI video is shifting from “cool demo” to “production tool.” The models competing for that crown — Google's Veo, Runway, Kling and now Vidu — are all racing on the same axes: consistency, resolution and speed. Vidu's angle is performances and references: fifteen image references and voice consistency attack exactly the pain point that breaks most AI-generated short dramas today, where a character's face or voice drifts mid-scene. Strong Chinese-text rendering inside frames, noted by early testers, is another differentiator for the short drama market where Vidu already has real users.
The timing is interesting too. Shengshu unveiled its real-time interactive S1 model in July, and Q4 Preview is the flagship offline-generation line pushing into cinematic territory. China's video models are increasingly trading blows with US labs on quality, not just price.
My take
Character consistency has been AI video's Achilles heel. If Q4 Preview's reference system delivers at the promised 4K/10-bit quality, short drama producers and ad teams get a genuinely useful tool — not just a toy. The proof will be in what creators ship with it over the next month. At $0.014 a second, trying it is cheap enough that the bar for excuses is now on the floor.
More AI tools like this: aipost.tech
Comments
Post a Comment