Skip to content
Wan 3.0 Uncensored cover art

Wan 3.0 Uncensored

Wan 3.0 Uncensored is Alibaba's video model, hosted on Vellria with nothing to install: text to video and image to video, 480p to 1080p, clips of 2–30s with sound. Run it in Studio or over the REST API, from 15 credits / s (100 credits = $1.00).

VideoText to video · Image to videoAlibabafrom 15 credits / s
vellria-wan-3

Wan 3.0 is developed by Alibaba. Vellria provides access to the model; it does not own it. All trademarks belong to their respective owners.

Input

Form

0 / 7500

Reference

0–10 images. You can upload files or paste https URLs; the provider downloads the image from the URL.

Settings

On

No generation yet

No generation yet

Output

No generation yet
Wan 3.0 Uncensored cover art

Cover art

from 15 credits / s

Sign in to run

Your request will cost 75 credits.

About Wan 3.0

Also known as: Wan 3 · Alibaba Wan 3.0

Wan 3.0 is the current video generation of Alibaba's Wan family. Most of the search interest still sits on Wan 2.2 — the open release people run locally for NSFW video with ComfyUI workflows and LoRAs — and on Wan 2.5. Wan 3.0 is their successor, not the same checkpoint: a newer model from the same team, hosted here so there is nothing to install, no VRAM to budget and no node graph to repair. The 2.2 checkpoint itself is also on this site as Wan 2.2, for short silent clips.

It is built for long, unbroken takes. Clips run longer here than on most of the catalog, so a shot can play out without a cut, sound included. Write a long take as one continuous description rather than a list of beats: who moves, where the camera goes, what changes in the light, what we hear. The model handles a slow build better than a montage.

Try framing and movement at a lower resolution first, then run the same idea higher up; the result can shift on that jump, so check it before committing a long clip to the top step. If the shot already exists as an image, hand it over as the first frame and trim it to the clip's proportions beforehand; otherwise describe everything in text and set the proportions yourself.

For the final version of a decided shot, step up to Wan 3.0 Pro. If you design stills in the same family, Wan 2.7 Image makes opening frames that animate naturally here. For a different take on audio-rich scenes, compare Seedance 2.5.

When to pick Wan 3.0

Pick it when

  • The shot should play out as a long, unbroken take with a slow build, camera movement and sound, without a cut.
  • You used Wan 2.2 locally and now want the newer Wan generation, with sound and longer takes, hosted without LoRAs or node graphs.
  • You designed opening frames on Wan 2.7 Image and want them animated inside the same model family.

Pick another model when

  • The shot is decided and the clip will be watched full screen on its own; render the keeper on Wan 3.0 Pro.
  • Your clip has to continue an existing image closely rather than reinterpret it; MiniMax H3 Max stayed close to the supplied frame in our runs.
  • You want an audio-rich scene with a different take on sound; compare Seedance 2.5 on the same idea.

Prompt or reference

Start from a picture when you already have the shot: it becomes the opening frame and sets the clip's shape, so crop before you upload. Write the scene from scratch when you do not, and pick the frame shape yourself.

How to use Wan 3.0 on Vellria

  1. Prompt: up to 7,500 characters in the panel above.
  2. Cost: from 15 credits / s — see pricing.

Example prompts

  • One continuous take: an adult woman wakes in a sunlit cabin, stretches, pulls on a flannel shirt and walks barefoot onto the porch, the camera follows her from the bed to the doorway, birdsong and creaking floorboards.

    Try this prompt
  • From the uploaded still: the adult couple in the hot tub talk and laugh, steam rising, snow falling around them, the camera slowly circles the tub once.

    Try this prompt
  • Long take of an adult man swimming laps in an outdoor pool at night, underwater lights, the camera tracks alongside him, water and distant music.

    Try this prompt
  • An adult woman in a velvet robe walks through a dark art gallery, her footsteps echoing, pausing in front of each lit painting, slow steadicam behind her.

    Try this prompt

Tips

  • Write the whole take as one paragraph in chronological order.
  • Prototype at the lowest resolution; re-check the result after moving up a step.
  • If you used Wan 2.2 locally, drop LoRA trigger words; describe the look in plain language instead.
  • Use an opening still to lock a character's appearance for long clips.

Specifications

Resolution:480p720p1080p
Duration:2–30s
Aspect ratio (Text to video):Auto (chosen from prompt)16:94:31:13:49:16
Reference images:1 reference image

See also: uncensored AI video generator · NSFW video generation · NSFW image to video.

Pricing

ResolutionCredits
480p15 credits / s
720p30 credits / s
1080p60 credits / s

Wan 3.0 vs Wan 3.0 Pro and Wan 2.2

Wan 3.0 is the standard tier of Alibaba's newest Wan video generation: long single takes with sound, from a prompt or a start frame. Wan 3.0 Pro shares its limits at a higher price, and Wan 2.2 is the older, silent and lower-priced release.

ModelPriceMax resolutionDurationReference images
Wan 3.0 Uncensored This pagefrom 15 credits / s1080p2–30sUp to 1
Wan 3.0 Pro Uncensoredfrom 20 credits / s1080p2–30sUp to 1
Wan 2.2 Uncensoredfrom 8 credits / s720p5–8sUp to 1

Prices and limits come from the live model catalog.

Wan 3.0 or Wan 3.0 Pro?

  • Price: from 20 credits / s (Wan 3.0: from 15 credits / s)

Wan 3.0 Pro is the higher tier of the same generation, with identical resolutions, lengths and frame shapes, so a take staged here moves over without edits. Spend its higher price per second only on the clip you keep.

Wan 3.0 or Wan 2.2?

  • Price: from 8 credits / s (Wan 3.0: from 15 credits / s)
  • Max resolution: 720p (Wan 3.0: 1080p)
  • Duration: 5–8s (Wan 3.0: 2–30s)
  • Prompt length: 10,000 characters (Wan 3.0: 7,500 characters)
  • Settings: None (Wan 3.0: Audio)

Wan 2.2 is the open release many people know from ComfyUI. It is silent and limited to short clips, but it costs less per second, which makes it useful for testing wording and motion.

Wan 3.0 API: price and a first request

Wan 3.0 costs from 15 credits / s over the API, the same rate as in Studio. Model id vellria-wan-3 on POST /v1/generate/video.

curl https://vellria.com/v1/generate/video \
  -H "Authorization: Bearer vll_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "vellria-wan-3",
    "prompt": "…",
    "duration": 2,
    "resolution": "480p",
    "ratio": "16:9"
  }'

Parameters, polling and uploads: API documentation · NSFW video API.

FAQ

Does Wan 3.0 allow NSFW?

On Vellria, yes, for fictional adult content. Nudity and explicit scenes between adults are allowed. Minors are refused automatically and real people may not be animated without consent.

Is Wan 3.0 free?

Free to start: a new account created with Google gets 20 free credits, valid for 7 days (about 10 images on Z-Image Turbo Uncensored), and every signed-in account gets free daily uses of the NSFW prompt generator and the image to prompt tool. The gift can go toward Wan 3.0 at from 15 credits / s; beyond that every run is paid from prepaid credits at the rate this page shows before you start, and a run that fails at the provider is refunded automatically. Is there a free uncensored AI generator? covers the cheapest way to a first result.

Is Wan 3.0 the same as Wan 2.2?

No. Wan 3.0 is the newer Alibaba Wan model that follows Wan 2.2, 2.5 and 2.6, with sound and much longer takes. The 2.2 checkpoint itself is hosted here too as Wan 2.2, for cheaper short silent clips.

Do I need ComfyUI or LoRAs?

No. Wan 3.0 runs hosted here or through the API; the prompt and an optional opening image are the only inputs.

Does Wan 3.0 have audio?

Yes, clips include sound generated from the scene you describe.

Can Wan 3.0 animate my image?

Yes. Attach a still and it becomes the opening frame; see NSFW image to video for tips.