Shengshu launches Vidu Q4 Preview with 4K output

Shengshu Technology launched Vidu Q4 Preview on October 7, 2026, positioning the model as a next-generation AI video tool for professional production teams. The company says the preview is available now through the Vidu platform website and is aimed at large-scale use cases including short drama, advertising, social content and film production.

The launch also comes with a split set of published pricing details. Yahoo Finance’s paid press-release pickup says pricing starts at $0.045 per second at 540P and rises to $0.39 per second at 4K, with image-to-video and reference-to-video generation discounted 30% through November 30. A separate Unite.AI report repeated a different self-reported starting price of $0.014 per second, underscoring that the pricing cited in syndications is not fully consistent.

What does Vidu Q4 Preview add for creators?

Vidu Q4 Preview adds reference-voice support, which Shengshu says is intended to keep a character’s voice consistent while preserving natural emotional delivery. The launch materials also say the model improves realism and lighting, and offers stable camera moves and cuts. In a creator testimonial carried in the release, the model was described as standing out for character performances and visual consistency, with strong handling of explosion lighting, smoke, dust, and Chinese text rendering.

The company describes Shengshu as originating from Tsinghua University AI research labs, and the launch materials frame Q4 Preview as the latest version of its Vidu AI video generation family. Shengshu also says public reports show it has raised nearly RMB 6 billion, or about $890 million, since February 2026.

How does the model work?

The product page described by Unite.AI lists two modes: Image-to-Video and Reference-to-Video. Image-to-Video animates one uploaded image from a text prompt. Reference-to-Video combines one to 15 images, and up to three audio references, to keep characters and scene elements consistent across generations.

Shengshu says the model supports cinematic 4K video with 10-bit color, and the company says creators can combine up to 15 reference images for complex scenes and concepts. Unite.AI reported output options spanning 540p, 720p, 1080p, 2K and 4K, with a 16-second maximum video length. It also reported that outputs preserve the original aspect ratio of uploaded material.

The same product-page summary says Reference-to-Video runs from 1 to 16 seconds, while Image-to-Video runs from 3 to 16 seconds. On the API side, Unite.AI reported an image-to-video endpoint that accepts a start-frame image up to 50MB, prompts up to 20,000 characters, durations of 3 to 16 seconds, and resolutions from 540p to 4K. It also reported that the audio parameter defaults to true and that the documentation supports audio-video synchronization plus automatic camera switching.

Why is Shengshu pitching it to production teams?

Shengshu is targeting Vidu Q4 Preview at independent creators, small studios and production teams working on narrative and commercial projects. The company and the product-page summary both list intended scenarios such as advertising, social content and cinematic production, while Unite.AI also reported uses including advertising concept testing, e-commerce variations, and filmmaking tests of performances, storyboards and pacing.

The launch materials emphasize iteration speed and efficiency. Shengshu says Q4 Preview can produce up to five times the output for the same budget under comparable conditions, and Unite.AI attributed to CEO Yihang Luo a view that the efficiency gains should let creators try more shots or versions. The company also tied the product to iterative production workflows such as adjusting expressions, camera angles and opening sequences.

Shengshu says the preview is being released before the full Q4 model so it can gather user input for the final version. That makes the current launch both a commercial release and a feedback stage for the product line.

What should users take away from the launch?

For founders and AI builders, the launch is notable less for a single feature than for how Shengshu is packaging the model: high-resolution video, multiple reference images, optional audio references, and workflow tools that aim at production rather than casual generation. The company is also using pricing, discounts and API documentation to reach both web users and developers.

At the same time, the published pricing is inconsistent across syndications, with one outlet citing $0.045 per second at 540P and another citing $0.014 per second as a starting point. Shengshu says final pricing, supported resolutions, features and usage terms may vary by plan and region, so buyers will need to check the product page and plan terms directly before budgeting around the launch rates.