Unseen Stack

B-roll & Stock

Wan AI (Alibaba)

Alibaba Cloud’s Wan consumer studio (Wan 3.0) — text/image/reference-to-video with native audiovisual output up to ~30s, omni multimodal refs, and instruction video editing — for custom faceless B-roll and short social clips.

What it is

Wan AI is Alibaba’s consumer generative image/video product on wan.video (create surface create.wan.video) — English global branding for the Wan / Tongyi Wanxiang line, powered by Alibaba Cloud. The homepage reviewed 2026-09-24 pushes Wan 3.0 with Native 30s, cinematic realism / immersive audiovisual, Omni-Creation, and precision video editing.

The faceless B-roll job is generative coverage: text-to-video, image-to-video, Reference to Video, or instruction-based Video Editing, export a short clip (marketing: up to about 30 seconds with native audiovisual), then finish voiceover, captions, and pacing in CapCut, VEED, Descript, Clipchamp, Kapwing, or Submagic. Official Features page (wan.video/features) lists Reference to Video (multi-person / human-object consistency with appearance and voice), Image to Video (up to 30s with auto scene splitting and native audio-visual framing), Text to Video, Video Editing, Text to Image, and Image Edit. Omni-Creation copy on Features: reference up to 10 images, 5 videos, and 5 audio tracks in one generation, plus document/webpage parsing and web-search grounding. Create UI chrome also names modes such as Reference to Video, First&Last Frame, Video Edit, Image Generation, and Agent (availability moves — trust the signed-in generate picker).

It is not Kling, PixVerse, Dreamina, Runway, Vidu, Luma, Hailuo, Higgsfield, Pollo, or DeeVid (already listed): those are rival first-party labs or multi-model desks. Wan is Alibaba’s native home for Wan 3.0 (some multi-model desks host a “Wan” line item inside their credit wallet — different seat). Prefer Kling when VIDEO Multi-Shot / native-audio membership math already wins; prefer Runway for suite depth; prefer PixVerse / Dreamina / Vidu / Hailuo / Luma when those looks and prices fit; prefer Pollo/DeeVid when you only want a thin multi-model desk that may already rent Wan-named models. It is not HeyGen / Synthesia (avatar presenters). It is not InVideo / Fliki / Pictory (script-to-stock assemblers). It is not CapCut / VEED / Descript / Clipchamp / Kapwing (timeline editors). It is not ElevenLabs as a standalone voice factory, Submagic / Opus Clip, Epidemic Sound / Artlist / Suno, Leonardo / Firefly / Midjourney / Canva, or Pexels / Pixabay.

Official framing: “Unleash Your Creativity,” Wan 3.0 longer narratives / greater control, open-source and API Platform footers for builders, and commercial-creation language aimed at production workflows. Operator entity on Terms: Intelligent Cloud Computing (Singapore) Private Limited; contact shown on site: wan_ai@service.alibaba.com.

Who it’s for

Use it if your bottleneck is custom generative B-roll with longer native-audio clips and multi-reference consistency — explainers, product lifestyle motion, character/prop-locked shorts, or 9:16/smart-ratio social clips — and you specifically want Alibaba Wan 3.0 rather than renting Wan-named models inside another lab’s wallet. Faceless fits: still or prompt → short Wan clip under an ElevenLabs/CapCut voiceover; Reference to Video for subject lock; Omni refs when you need image+video+audio grounding before timeline polish.

Skip Free as a monetized publish plan — Pro/Premium cards explicitly sell watermark-free downloads; Free does not. Skip if Kling, Runway, PixVerse, Dreamina, Vidu, Luma, Hailuo, Pollo, or DeeVid already owns your generative seat and consolidating would not cancel spend. Skip treating unused monthly membership credits as a bank — official Terms: membership Credits are valid one month from distribution unless the live page says otherwise. Skip if you only need free filmed stock (Pexels/Pixabay) or a presenter avatar as the whole product (HeyGen/Synthesia). Skip expecting a polished NLE — finish cuts, captions, and Safe Zones elsewhere. Skip assuming create.wan.video credits buy DashScope/API access — pricing page says those are separate purchases.

Standout details

  • Official surfaces: wan.video · Features wan.video/features · Create create.wan.video · Pricing create.wan.video/pricing · Terms wan.video/policy/termsofService · Usage Policy wan.video/policy/usagePolicy · footer also points to Open Source / GitHub, API Platform, Privacy Policy, and Training Data Summary.
  • Product names in market: Wan, Wan AI, Wan 3.0, Tongyi Wanxiang lineage; company framing: Alibaba / Alibaba Cloud.
  • Wan 3.0 capabilities (homepage + Features, 2026-09-24): T2V · I2V · Reference to Video · instruction Video Editing · Omni multimodal refs (up to 10 images / 5 videos / 5 audio) · native audiovisual / binaural marketing · up to ~30s · High-Res 1080p on paid cards · create UI also exposes 720P / Smart Ratio / duration controls.
  • Consumer plan economics: Pro at US$5/mo yearly-effective (300 credits) is the clear watermark-free entry on the live yearly view; Premium US$20/mo yearly-effective (1,200 credits) for higher concurrency and credit headroom. Credit burn is model/duration/resolution dependent — trust the in-app estimate before you batch. Accelerate marketing on Pro equates 300 credits to “up to 60 videos or 1,200 images” (and Premium 1,200 → 240 videos / 4,800 images) — treat as upper-bound marketing, not a guarantee for long 1080p Omni runs.
  • Rights / watermark: Pro/Premium pricing cards — “Download watermark-free images & videos.” Terms §II — do not remove/obscure labels or watermarks on Outputs without proper permission. Terms §III.4 — as between you and Wan, you retain rights in Prompts; subject to compliance, Wan assigns its right/title/interest (if any) in Outputs generated at your request. Usage Policy covers both non-commercial and commercial use of platforms/APIs/open-source models under its safety rules — still clear third-party likeness/IP in prompts and read live Terms before monetized publish. Wan also takes a broad license to use User Content to provide/improve services (Terms §III).
  • Credits: membership Credits monthly, typically 1-month validity (Terms VI.7); purchased credit packs 2-year validity, non-refundable/non-transferable (pricing Credits tab).
  • Traffic / search-intent justification: AICPB product page for Wan AI - 阿里 (last updated Jul 8, 2026) cites #14 · ~2.44M website visits in AI Video Generators Rankings by Website Visits for Jun 2026 — the strongest remaining unlisted name on that public table after Hailuo shipped, ahead of still-unlisted Vidnoz (~2.28M) and Pika Art (~2.04M) on the same Scout shortlist. Treat third-party visit estimates as directional, not audited.

Honest verdict

Try it. After Hailuo, Wan/Alibaba is the clearest next Unseen Stack ship in the generative B-roll lane on public AICPB traffic — a first-party Wan 3.0 lab with native audiovisual, Omni multi-ref, ~30s marketing length, and a paid watermark-off path starting at Pro’s yearly-effective list price — but Kling, PixVerse, Dreamina, Runway, Vidu, Luma, Hailuo, Pollo, and DeeVid already cover a lot of the same faceless job.

Start on Free / daily check-in credits only to taste Wan 3.0 prompts, Reference/Omni workflows, and credit quotes. Move to Pro (or confirm the live entry commercial + watermark-free card) before monetized Shorts/TikToks, and watch long 1080p / Omni burns against the 300-credit Pro wallet. Prefer Kling when Multi-Shot / starter-membership math already fits; prefer Runway for suite depth; prefer PixVerse / Dreamina / Vidu / Hailuo / Luma when those looks win; prefer Pollo/DeeVid when one thin multi-model desk is enough (and may already expose a Wan row); prefer Wan when Alibaba-native Wan 3.0 length, omni refs, and Pro’s list price are the fit. Re-check create.wan.video/pricing stickers (yearly vs monthly vs gift-card US$10/US$40), generate-panel credit quotes, watermark path, Terms vs live cards, and the create-vs-API wallet split the day you subscribe — plan names and entitlements move, and unused membership credits do not behave like a long-term bank.