RefVid

Wan 3.0 AI Video Generator

Turn your own photos into videos in minutes · Your first video is free, no watermark

  • Free first video
  • Wan 3.0
  • No watermark
  • Commercial use
  • Unboxing video of LUMEN, a serum bottle with a dropper
  • Product 360 video of AERO 1, a grey and orange running shoe
  • Example video of the Wallet Open and Close template
  • Floating Product video of AERO 1, a grey and orange running shoe
  • Bullet Time video of LUMEN, a serum bottle with a dropper
  • Water Splash video of LUMEN, a serum bottle with a dropper
  • UGC Ad video of Mia and LUMEN
  • Flat Lay video of EMBER, a black bag of coffee

First video free, no card

Wan 3.0 is Alibaba's video model, officially launched on August 24, 2026. It makes clips of up to 30 seconds, twice Wan 2.7's 15, from text, images, video and audio references.

On RefVid it turns 1 to 10 photos of your product or character into a 3- to 30-second clip at 720p or 1080p, with sound on or off, or makes a clip from a prompt alone. The box above is already set to Wan 3.0.

Wan 3.0 on RefVid: specs

SpecWan 3.0 on RefVid
Clip length3–30 seconds
Resolution720p or 1080p
Shapes16:9, 9:16, or 1:1
SoundOn or off, same price
Reference photos1–10 photos of one subject
Video from a prompt aloneYes
End frame (a picture the clip ends on)No
Photos of real-looking peopleYes
Longest prompt19,700 characters
Credits per second, 720p37, with or without sound
Credits per second, 1080p73, with or without sound
From RefVid's own settings for Wan 3.0, checked 2026-11-06. The studio shows what a clip costs before it runs.

What Wan 3.0 is

  • Wan 3.0 is the third generation of Alibaba's Wan video models. Alibaba announced a beta on August 7, 2026 and the official launch on August 24. Its API makes 2 to 30 seconds at 480p, 720p or 1080p and takes up to 10 reference images, 5 reference videos and 5 audio clips, plus a first and last frame, according to Alibaba Cloud's documentation.
  • RefVid runs Wan 3.0 through fal.ai's reference-to-video endpoint. In the RefVid box you add up to 10 photos of one subject, pick a length from 3 to 30 seconds, 16:9, 9:16 or 1:1, 720p or 1080p, and switch sound on or off; sound costs nothing extra on Wan 3.0. With no photos, it makes a clip from your prompt alone.

Wan 3.0 vs 2.7: what changed

  • Alibaba says Wan 3.0 roughly doubles the longest clip, from Wan 2.7's 15 seconds to 30. References grow from 5 images and videos combined, plus one reference voice, to 10 images, 5 videos and 5 audio clips, and Wan 3.0 adds a sound switch and 480p. On fal, 720p stays at $0.10 a second.
Wan 2.7Wan 3.0
ReleasedApril 7, 2026August 24, 2026 (beta August 7)
Clip length2–15 seconds (2–10 with a reference video)2–30 seconds
Resolutions720p, 1080p480p, 720p, 1080p
Shapes16:9, 9:16, 1:1, 4:3, 3:421:9, 16:9, 4:3, 1:1, 3:4, 9:16 or adaptive
ReferencesUp to 5 images and videos combined, plus a 1–10 s reference voiceUp to 10 images, 5 videos and 5 audio clips
SoundNo switch publishedOn or off, on by default
Price per second on fal$0.10$0.10 at 720p, $0.20 at 1080p
Sources: Alibaba Cloud blog posts and Model Studio API references for Wan 2.7 and Wan 3.0, and fal.ai's Wan 2.7 and Wan 3.0 reference-to-video pages, read November 4, 2026.

Wan 3.0 reference to video: long clips, same subject

  • On RefVid the photos you add are saved as one subject, so every Wan 3.0 clip starts from the same product or character. Its 30-second limit suits clips that need time: an unboxing from sealed box to product in hand, a walk through a room, a short scene with several actions in a row.
  • Give it 3 to 5 photos from different angles in plain light, then write the actions in order, for example: "Hands cut the tape, lift the lid, take the headphones out and turn them toward the camera; the camera stays close and follows the hands." One action per sentence keeps a long clip in order.

Make a Wan 3.0 video in three steps

  • Add 1 to 10 photos of your product or character in the box at the top of this page, or start from a template in the row above it. The box is already set to Wan 3.0 and 720p.
  • Write what should happen, in order. Settings changes the length (3 to 30 seconds), the shape (16:9, 9:16 or 1:1), 1080p and sound.
  • Press Generate. If you are not signed in, you sign in first and the clip starts right after. A clip usually takes 2 to 10 minutes, and we email you when it is ready.

What a Wan 3.0 clip costs on RefVid

  • Every clip is priced in credits per second, as in the table above, with sound at the same price, and the studio shows the total before anything runs. New accounts get free credits for a first clip, with no watermark. After that, single videos start at $2.99, one-time credit packs at $9.90 and monthly plans at $10. A failed clip refunds its credits automatically.

Wan 3.0 next to every RefVid model: measured clips, wait and cost

  • These numbers are our own, not the vendor's: every clip people outside our team started on RefVid from October 27 – November 6, 2026, split by the model they picked, with how many finished and the median wait from the start of the generation to the finished clip. Next to them are each model's photo limit, clip length and what a 5-second clip costs today, so Wan 3.0 can be read against the other models in one table.
  • The finished share and the wait are shown only where a model had at least 20 finished or failed clips in the period; the models added most recently have fewer so far. A failed clip costs nothing: its credits go back to the account. The table follows the newest snapshot of our usage statistics.
ModelClips (finished or failed)FinishedMedian waitPhotos of one subjectClip lengthCredits for a 5-second clipCost with the $9.90 pack
Kling 418Under 20 clipsUnder 20 clips103–30 sec290$1.91
Kling 3.0 Omni31083%2.1 min103–15 sec155$1.02
Seedance 2.53Under 20 clipsUnder 20 clips104–30 sec860$5.68
Wan 3.013Under 20 clipsUnder 20 clips103–30 sec185$1.22
Vidu Q42171%4.2 min103–16 sec125$0.83
Gemini Omni Flash1Under 20 clipsUnder 20 clips103–10 sec185$1.22
PixVerse C18099%1.4 min73–15 sec95$0.63
Happy Horse 1.11Under 20 clipsUnder 20 clips93–15 sec255$1.68
MiniMax H31Under 20 clipsUnder 20 clips55–15 sec110$0.73
Source: RefVid database, every clip started from October 27 – November 6, 2026, our own and test accounts left out; snapshot of November 6, 2026. Photos and clip length are RefVid's settings for each model. Credits are for a 5-second clip at the model's standard resolution, without sound unless the model always adds it (Kling 4); the cost uses the $9.90 credit pack.

When another model fits better

  • Wan 3.0 suits long clips and clips from words alone. For short clips at a lower price per second, Vidu Q4 and PixVerse C1 cost less. Kling 4 also reaches 30 seconds and always has sound. Your saved subjects work with every model, so you can try the same photos on two of them.

Start from a template

Each one opens the studio with the shots, shape and length already set. Add your own product and generate.

More on AI video models

Text to video AI →

Frequently asked questions

Is Wan 3.0 free on RefVid?

New accounts get 180 free credits for a first clip, with no watermark and no card. A Wan 3.0 clip costs a little more per second than the free clip's model; if a clip costs more than your credits, the box says so before anything runs and offers what fits.

How much does Wan 3.0 cost?

On fal.ai, Wan 3.0 reference to video costs $0.10 a second at 720p and $0.20 at 1080p (read November 4, 2026). On RefVid a clip is priced in credits per second, with or without sound at the same price; the table above has the numbers.

How long can a Wan 3.0 video be?

Up to 30 seconds in one generation, according to Alibaba, twice Wan 2.7's 15. On RefVid you pick 3 to 30 seconds in Settings.

Can Wan 3.0 make a video from text alone?

Yes. On RefVid, Wan 3.0 makes a clip from your prompt alone when you add no photos, at the same price per second as a clip from photos.

How many reference images does Wan 3.0 take?

Up to 10, according to Alibaba Cloud's API reference, plus up to 5 reference videos and 5 audio clips. On RefVid you add up to 10 photos of one subject.

By All guides