AlibabaWan 3.0 · by Alibaba

    Wan 3.0 AI Video Generator — one take, 30 seconds

    Wan 3.0 is the newest generation of Alibaba's open-weights video family: one continuous take up to 30 seconds with audio, prompts up to 20,000 characters, and the same unfiltered hosting as the rest of the Wan line. Run it online on OpenGradient without a GPU or API keys — type a prompt, pick a format, render.

    Private, anonymous AIOne take up to 30sUncensored open weights
    Start creating free

    Free credits to start · no card · lands with Wan 3.0 preselected

    Real clips generated on OpenGradient

    Capabilities

    What you can make with Wan 3.0

    The unrestricted lane's flagship: half-minute single takes with sound, from prompts as long as the brief needs.

    30 seconds in one take

    Durations from 5 to 30 seconds rendered as one continuous take — matched only by Seedance 2.5 — with room for a scene that has a beginning, a middle and an end.

    Unrestricted prompting

    Wan 3.0 runs as an open-weights deployment with no provider-side prompt moderation — briefs that hosted models decline run here, within the published content rules.

    Audio with the picture

    Sound is generated in the same pass as the 1080p frames, so even a 30-second take arrives watchable — no separate audio step, no silent drafts.

    Reference images & last frame

    Image-to-video takes up to 10 reference images for character and product consistency, plus last-frame control for chaining takes into longer sequences.

    Wan 3.0 at a glance

    Modes

    Text to video + Image to video

    Clip length

    5s – 30s

    Aspect ratios

    16:9 · 9:16 · 1:1 · 4:3 · 3:4

    Audio

    Native audio

    Moderation

    None — open weights, 18+ gate

    Variants

    8 live on OpenGradient

    The Wan lineup, live on OpenGradient

    Wan 3.0 · text-to-video · NewWan 2.7 · text-to-video · UncensoredWan 2.6 · text-to-video · UncensoredWan 2.5 · text-to-video · UncensoredWan 2.2 14B · text-to-video · UncensoredWan 3.0 · image-to-video · NewWan 2.6 · image-to-video · UncensoredWan 2.2 14B · image-to-video · Uncensored

    Where it fits

    How people use Wan 3.0

    The model for long, unrestricted takes — one render instead of a stitched timeline.

    Whole scenes in one render

    A 30-second beat — setup, turn, payoff — generated as a single take instead of assembled from 5-second fragments.

    Briefs everything else bounces

    Mature themes, edgy satire, horror — prompts that trip other providers' filters run without the permission dance.

    Production-brief prompts

    Prompts run to 20,000 characters — the longest in the catalog — so a full shot list with camera, lighting and audio direction fits in one prompt.

    Cheap iteration loops

    Draft on Wan 2.5 for cents, refine the prompt, then final-render the keeper as one long take on 3.0.

    Questions, answered

    Frequently asked

    What is Wan 3.0?
    Wan 3.0 is the newest generation of Alibaba's open-weights Wan video family. It renders one continuous take up to 30 seconds with audio from a text prompt or an image, and runs on OpenGradient with no provider-side prompt moderation.
    What's new compared to Wan 2.7?
    Clip length triples: one take up to 30 seconds instead of 10. Prompts grow to 20,000 characters, and image-to-video adds up to 10 reference images plus last-frame control for chaining takes. Audio and the unfiltered hosting carry over.
    Is Wan 3.0 really uncensored?
    It's hosted as an open-weights deployment with no provider-side prompt moderation, like the rest of the Wan family. That's not the same as no rules: illegal content is banned, and explicit capabilities sit behind an 18+ gate.
    How do I run Wan 3.0 without a GPU?
    You don't need one. Open OpenGradient's Video Studio from the button on this page — Wan 3.0 arrives preselected. Type a prompt, pick an aspect ratio and duration, and render. Same weights as self-hosting, none of the plumbing.
    What does a clip cost?
    Per-clip prices depend on the model and duration and are shown before you generate. 1,000 credits = $1; packs start at $10; you get free credits to start - no card needed.
    Wan 3.0 or Seedance 2.5?
    They're the only two models here that render 30 seconds in one take. Seedance 2.5 leads on overall polish and synced dialogue but is provider-filtered; Wan 3.0 is the unrestricted pick and takes the longest prompts in the catalog. They share one picker, so audition your brief on both.
    FREE TO START • NO CARD

    Render a 30-second take on Wan 3.0

    No GPU, no API keys — free credits to start. Type a prompt and watch one continuous take arrive in minutes.

    Unrestricted means no provider-side prompt moderation - not no rules. Illegal content is banned, explicit capabilities are gated 18+, and sexual content depicting real people without their consent is prohibited.