Vidu Q4 Preview sells AI video from $0.014 a second at launch

Shengshu Technology launched Vidu Q4 Preview on October 7 with launch pricing advertised from $0.014 a second. That floor is real, and it is promotional. Shengshu's Beijing release and Vidu's price list put the list price at $0.045 a second at 540p, climbing to $0.39 a second at 4K.

Vidu Q4 Preview is the first public preview of Shengshu's next-generation flagship model for expressive audio and video. Shengshu says it coordinates facial expression, emotion, body movement and voice, holds camera moves and cuts together through fights and chases, and seats effects such as explosions and particles inside the scene. The model takes up to 15 reference images and 3 reference audio clips, and outputs at 540p, 720p, 1080p, 2K or 4K, with 10-bit color at the top tiers. Audio is on by default.

The price is the story. Per Vidu's pricing page, Q4 Preview bills at 9 credits a second at 540p, 19 at 720p, 24 at 1080p, 38 at 2K and 78 at 4K, at a standard rate of $0.005 a credit. That is $0.045 a second at 540p and $0.12 a second at 1080p. Image-to-video and reference-to-video run at a 30% discount through November 30, and off-peak jobs bill at half. Unite.AI carried the $0.014 headline without the list price beside it.

Set against the rivals on their own pages, the launch rate leads the field while the list rate lands mid-pack:

How Vidu Q4 Preview's price compares

Wan 3.0's lowest rate is the $0.041256 a second Alibaba lists for its China, Japan, Germany, US and Hong Kong regions; its Singapore region is $0.05. The figures come from Alibaba's Wan 3.0 page, MiniMax's pay-as-you-go page and BytePlus's ModelArk price list.

The independent signal points the same way. OrcaRouter, reading the Artificial Analysis leaderboards on October 8, reports Vidu Q4 Preview at Elo 1,179 on the AA-Video-I2V v1.0 board, third of the models listed and within a point of Gemini Omni Flash. Two MiniMax models sit above it. The same write-up records the model's measured rate at $7.20 a minute, which is the 1080p list price. Vidu Q3 Pro, the model it replaces, scored 1,056 in nineteenth, a 123-point step in a single release.

One gap matters as much as the rank. Vidu Q4 Preview does not appear on the text-to-video board, because it has no text-to-video mode. Vidu's product page lists image-to-video and reference-to-video only, and BlockBeats makes the same point. Clips run 3 to 16 seconds for both image-to-video and reference-to-video, and the API returns a file that stays live for 24 hours.

Shengshu's price direction runs one way. The Beijing company, born out of Tsinghua University's AI research, shipped its real-time Vidu S2 avatar and editing models on September 15 and Vidu Q3 before that.

What to watch is the full Q4 model. Shengshu is selling the preview in production now, but the maker says the release exists to shape the final version, which is when a promotion like this one tends to end.

Vidu Q4 Preview is live now through Vidu's web product and API at platform.vidu.com; sign-up comes with trial credits, and the image-to-video and reference-to-video promo rate runs through November 30.