Vidu AI 影片生成

Vidu Q4 Preview is ShengShu's newest model, available on FlyAIgh as a preview. It has two modes. Image-to-video takes one still as the first frame and keeps that image's aspect ratio and look. Reference-to-video composes up to 12 reference images — characters, scenes, props — into one shot in 16:9, 9:16, 1:1, 4:3 or 3:4. Clips run any whole number of seconds from 3 to 16, at 540p, 720p, 1080p, 2K or 4K, and dialogue and sound effects are generated together with the picture. Q4 has no text-to-video and no first/last-frame mode; Q3 Pro covers those.

On FlyAIgh, Vidu Q4 Preview costs 6 credits per second at 720p — 30 credits for a 5-second clip. A 5-second clip is 10 credits at 540p, 37 at 1080p, 60 at 2K and 120 at 4K. Sound is included at no extra cost.

Vidu Q3 Pro is ShengShu Technology's third-generation flagship video model. ShengShu is a Beijing-based AI research lab spun out of Tsinghua University, and Vidu has built a reputation for strong reference-driven generation and unusually consistent motion physics — particularly useful for character animation and product-led shots.

The Pro variant ships with full mode coverage: text-to-video, image-to-video (single first frame), first/last frame control (two reference images bookending the clip), and reference-to-video using up to 7 reference images. This is the same reference budget as Kling V3 Omni and more than HappyHorse's auto-routing mode.

Unlike most video models on FlyAIgh, Vidu Q3 Pro uses fixed durations rather than continuous picks — clips are exactly 4, 5 or 8 seconds. The trade-off is that each duration is highly optimized for that specific length, giving better motion coherence than models that span a wide duration range. Output is 720p in 16:9, 9:16 or 1:1 aspect.

Pricing is 8 credits per second — 40 credits for a 5-second clip, the same per-second rate as Kling V3. The 5-second clip is the most common pick for social-format content; 8-second is the right choice when you need a beat of action plus a held final frame.

Vidu is particularly strong at character motion, multi-reference composition (drop in multiple subjects and have them coexist in the same shot) and stylized illustration. It is weaker than Sora 2 Pro for photorealistic faces and weaker than VEO 3.1 Quality for ultra-fine textural detail. Pick Vidu when you have multiple reference images and want strong subject consistency, or when fixed-duration motion coherence matters more than the longest possible clip.

There are other Vidu SKUs (Q3 Turbo, Q2) not yet exposed on FlyAIgh — contact support if your workflow needs them.