What is Vidu Q4 Preview?
Vidu Q4 Preview is the first public preview of Q4, the next-generation flagship video model from Shengshu Technology (生数科技), released on October 7, 2026. It focuses on lifelike character performance, cinematic camera language and complex effects, and generates video with synchronized native audio — dialogue and sound effects produced together with the visuals. The final Q4 model has no announced release date; the preview exists so creators can apply it to real projects and feed results back into the final build.
Capabilities
- Image-to-Video — one starting image plus a prompt, 3–16 seconds per clip.
- Reference-to-Video — up to 15 image references to lock characters, wardrobe, props, products and environments, plus up to 3 reference audio clips to keep a character's voice consistent across a scene. 1–16 seconds.
- Native audio — synchronized dialogue and sound effects when audio output is enabled.
- Up to 4K — 540p, 720p, 1080p, 2K and 4K output with 10-bit color at higher tiers, 24fps.
- Note: the preview exposes image-to-video and reference-to-video only; no text-to-video or first-and-last-frame endpoints are documented yet.
Pricing
Launch pricing is genuinely confusing: Shengshu's two same-day press releases quote different starting prices — $0.014/second in one and $0.045/second at 540p in the other (scaling to $0.39/second at 4K). A 30% launch discount runs through November 30, 2026. Official API pricing in RMB lists 0.03125 RMB per credit, which works out to roughly ¥0.75/second at 1080p and ¥2.44/second at 4K. Treat any cost math this month as provisional until a single rate card is published.
Benchmarks and Reception
Third-party tracking places Vidu Q4 Preview third on Artificial Analysis's image-to-video leaderboard with an Elo of 1,179, well ahead of the older Vidu Q3 Pro (1,056). CEO Yihang Luo has framed the release as a push for models to "prove themselves beyond carefully selected demos."
How to Use It
Vidu Q4 Preview is available through Vidu's web product and the Vidu API platform (model ID viduq4-preview). Developers should start from the official image-to-video (/ent/v2/img2video) and reference-to-video (/ent/v2/reference2video) endpoints.
How It Compares
- Seedance 2.5 (ByteDance) — longer native clips (30s with extensions), up to 50 references, strong audio-video joint generation. Vidu Q4 Preview counters with 15-image + 3-audio references, 4K 10-bit output and aggressive launch pricing.
- Kling — strong motion quality and mature tooling; Vidu's differentiator is multi-reference character/voice consistency at 4K.
Frequently Asked Questions
Is Vidu Q4 Preview free?
Generation is pay-per-use, but the launch carries a 30% promotional discount through November 30, 2026. Check the official site for current credit pricing and any free trial quota.
What is the API model ID?
viduq4-preview. Do not substitute guessed IDs such as vidu-q4 or viduq4.
Can it generate video from text only?
Not in the preview — the documented endpoints are image-to-video and reference-to-video. Start from an image.
Findry coverage
- Vidu Q4 Preview: pricing, features, and comparisons — the two official launch prices explained, plus Seedance/Kling context
