Vidu AI

What it is

Vidu AI (vidu.com) is Shengshu Technology’s first-party generative video lab for short clips from text, images, and multi-image references. The faceless B-roll job is generative coverage: type a prompt or drop a still (or a small reference set), pick a model/mode, get a short MP4, then finish captions and pacing in CapCut, VEED, Descript, Clipchamp, Kapwing, or Submagic.

Official homepage FAQ and product surfaces name Text to Video, Image to Video (including first- and last-frame control), and Reference to Video (upload three or more images of a character, object, or scene; My References for reuse). Marketing and creator testimonials lean hard on 2D/anime motion, multi-reference consistency, Storyboard-style workflows, and Off-Peak Mode. The flagship consumer/API tier called out in official docs is the Vidu Q3 series (pro / turbo / mix / drama / ad variants on the platform model map): up to 1080p, 24 fps, durations up to 1–16 seconds on Q3 text/image paths, with optional native audio-video output (dialogue / SFX) on supported Q3 models. Older Q2 / Q1 / Vidu 2.0 rows remain on the API model map for shorter or cheaper generations. A separate developer Open Platform lives at platform.vidu.com (API keys, credit top-ups, debug console). API Terms of Use are with ShengShu AI HK Limited (Hong Kong law / HKIAC arbitration); consumer support is support@vidu.com; Discord is linked from the homepage FAQ.

It is not Kling, PixVerse, Runway, Dreamina, or Higgsfield (already listed): those are rival first-party (or CapCut-family) labs — Vidu is another native model home, not a wrapper. Prefer Kling when Multi-Shot / Omni / longer native-audio story already owns your pipeline; prefer PixVerse or Dreamina when those labs’ look and credit math already win your niche; prefer Runway when Gen-4.x / Aleph suite tools matter more than Vidu’s reference-to-video + anime specialization. It is not DeeVid or Pollo (already listed): those are multi-model credit desks that may host a Vidu-named model — this listing is the first-party Vidu.com product. It is not HeyGen / Synthesia (avatar presenters as the product). It is not InVideo / Fliki / Pictory (script-to-stock assemblers). It is not CapCut / VEED / Descript / Clipchamp / Kapwing (timeline editors). It is not ElevenLabs (voice factory), Submagic / Opus Clip (caption or long-to-shorts specialists), Epidemic Sound / Artlist / Suno (music), Leonardo / Firefly / Midjourney / Canva (still-first desks), or Pexels / Pixabay (filmed stock).

Homepage and pricing marketing claim millions of users across 200+ countries — treat as vendor marketing, not an independent audit.

Who it’s for

Use it if your bottleneck is custom motion stock cannot match — especially character- or product-consistent shorts, anime/2D-leaning faceless channels, vertical TikTok/Shorts B-roll, or reference-led product motion — and you want a first-party lab with published Reference-to-Video + Q3 native-audio options rather than only renting “Vidu” inside a multi-model desk. Faceless fits: prompt or image-to-video coverage under a voiceover, recurring mascot/character shots via My References, and Off-Peak experiments when you can wait for cheaper queue time. Pay at least Standard (or whatever the live paid card is) so exports lose the watermark and commercial-publishing language matches what you publish.

Skip Free as a publish plan — official blog posts state free output is watermarked and personal/non-commercial only. Skip if Kling, PixVerse, Dreamina, Runway, Pollo, or DeeVid already owns your generative seat and consolidating would not cancel spend. Skip expecting a full NLE, claim-safe music library, or 4K masters as the default consumer ceiling (official surfaces advertise up to 1080p for Q3 video; API Upscale Pro is a separate paid path). Skip if you need refunds after a bad month — official pricing FAQ says Vidu cannot provide refunds. Skip treating “~N videos” marketing as a guaranteed clip quota — burn scales with model, duration, resolution, and audio. Skip if you need filmed documentary stock (Pexels/Pixabay) or a presenter avatar as the whole product (HeyGen/Synthesia).

Standout details

• Official surfaces: vidu.com · pricing vidu.com/pricing · help vidu.com/help-center · support support@vidu.com · Discord linked from homepage FAQ · developer platform platform.vidu.com · API docs/pricing platform.vidu.com/docs/pricing · API Terms platform.vidu.com/docs/terms-of-use (effective 2025-01-23; ShengShu AI HK Limited).
• Modes named on official homepage FAQ: Text to Video · Image to Video (first/last frame) · Reference to Video (3+ images; My References library).
• Models (platform model map, reviewed 2026-09-20): ViduQ3 series (pro / mix / drama / ad / turbo) up to 1080p, 24fps, typically 1–16s (drama/ad rows 3–15s) with simultaneous audio & video on several Q3 variants · older Q2 / Q1 / Vidu 2.0 rows for shorter/cheaper work · image, audio, lip-sync, motion-control, extension, and one-click ad/MV tools on the same platform surface.
• Consumer pricing page (2026-09-20 fetch): marketing bullets (Ultra-Fast Generation, Multi-Reference Consistency, Off-Peak Mode, High-Quality Video, Smooth 2D Animation) + FAQs; dollar/credit plan cards are dynamically loaded — reconfirm in-browser or signed-in checkout. Official FAQ: Subscription credits expire in 30 days; purchased/bonus credits last 2 years; Stripe auto-renew; no refunds.
• Watermark / commercial (official blog, not just third parties): Free = watermark + personal/non-commercial; paid = watermark removed + commercial publishing; Standard framed as the entry paid tier for monetized TikTok (vidu.com/blog posts May–Jun 2026). API Terms separately say the Open Platform license does not restrict commercial purposes (subject to prohibited-use rules) — that is the API surface, not a substitute for consumer plan terms.
• API credit stickers (platform.vidu.com/docs/pricing, 2026-09-20): US$0.005 per credit; example Q3-pro img/text/start-end 24 credits/s @1080p or 20/s @720p; Off-Peak halves many rows; reference-to-video and Q2/Q1 tables differ — budget retries.
• Off-Peak: official API docs — lower credit cost; tasks may take up to 48 hours; unfinished off-peak tasks can cancel with credit return. Consumer Off-Peak Mode is marketed on the pricing page; confirm eligibility per plan in-app.
• Traffic / search-intent justification: AICPB AI Video Generators Rankings by Website Visits — Issue 37 (Jun 2026), last updated Jul 8, 2026, lists Vidu (~3.17M, #10) ahead of still-unlisted Luma AI (~2.86M), Hailuo AI (~2.76M), and Pika Art (~2.04M), and behind already-listed Dreamina, Kling, HeyGen, Pollo, PixVerse, Runwayml, and DeeVid. After DeeVid shipped, Vidu is the strongest remaining unlisted first-party generative B-roll lab on that public AICPB table. (Separate June 2026 OneLittleWeb estimate puts Hailuo ~4.87M ahead of Vidu ~4.07M — noted, but Stack Scout keeps AICPB as the lane’s consistent ranking source unless a later AICPB issue flips the order.)

Honest verdict

Try it. On the public AICPB table Vidu is the clear next Unseen Stack ship in the generative B-roll lane after DeeVid — a first-party lab with Reference-to-Video, Q3 native audio, and Off-Peak experimentation — but Kling, PixVerse, Dreamina, and Runway already cover a lot of the same faceless job, and Pollo/DeeVid may already expose Vidu-named models inside a multi-model wallet.

Start on free credits / Off-Peak only to taste prompts, reference consistency, and watermarked exports. Move to Standard (or the live entry paid card) before you publish monetized Shorts/TikToks so the watermark is gone and commercial language matches paid-plan marketing. Prefer Kling when Multi-Shot / Omni already fits; prefer PixVerse or Dreamina when those looks win your niche; prefer Runway for suite depth; prefer Pollo/DeeVid when one multi-model desk is enough; prefer Vidu when first-party reference consistency, anime/2D strength, Q3 audio, and this credit math beat the alternatives. Re-check vidu.com/pricing stickers (after the JS table loads), generate-panel credit quotes, watermark path, blog commercial notes, and Terms the day you subscribe — models, promos, and entitlements move, and refunds are explicitly off the table.