Up to 30 seconds in one pass
Wan 3.0 doubled the longest clip of the versions before it. A recipe, a review or a coaching cue can run as one continuous take.
Wan · by Alibaba
Wan AI video generator
Wan is Alibaba's video model family. Wan 3.0 runs up to 30 seconds in a single generation, with dialogue and sound made alongside the picture, so a tutorial, a review or a walking vlog can play out start to finish without cuts. Write a scene below and it opens in Video Studio on Wan 3.0.
Example video
Opens in Video Studio with the model you picked. You see the price in credits before anything runs.
What Wan is known for
Sound with the picture
Dialogue, ambience and effects are generated together, at no extra cost. Every scripted line in our gallery was spoken as written.
Beats you can time
Write the scene as timecoded beats (0–7 s, 7–14 s…) and Wan follows the order. Our pasta clip hits all three beats in 20 seconds.
Exact length, any second
Pick any length from 2 to 30 seconds and pay for that, instead of rounding up to a fixed 5 or 10.
Where it's weaker
- Long clips are where small errors build up: check faces and hands near the end of a 20–30 second take.
- A spoken line can stutter or repeat: our headphone reviewer says "is unreal" twice.
- Wan rewrites short prompts to add detail; put every must-keep detail in the prompt yourself.
- Text and numbers on screen need checking frame by frame.
- Wan 3.0 is not open source; the last open-weights Wan is 2.2.
- Maker
- Alibaba · Wan 3.0 August 2026
- Clip length on Affogato
- Any length from 2 to 30 s (Wan 3.0)
- Resolution
- 480p · 720p · 1080p
- Aspect ratios
- 9:16 · 3:4 · 1:1 · 4:3 · 16:9
- Sound
- Native dialogue and effects, on by default
- Inputs
- Text · start and end frame · reference images, clips and audio
Made with Wan on Affogato
Every clip below was generated with the prompt shown, unedited. Load any prompt into the generator and make your own version.
20 s in one take
A pasta tutorial in three beats
Prompt
Vertical cooking tutorial in one continuous take, three beats. 0 to 7 seconds: a home cook in a linen apron drops fresh pasta into a pot of boiling water and says, "two minutes, no more." 7 to 14 seconds: she tosses the drained pasta in a pan with butter and lemon zest, steam rising. 14 to 20 seconds: she twirls a forkful, tastes it, closes her eyes and nods. Warm kitchen light. Sound: bubbling water, sizzling butter, her voice, no music.
UGC review
Headphone unboxing
Prompt
Vertical UGC product review filmed on a phone: a young man with a fade haircut and a hoodie sits on his bed, opens a box and pulls out a pair of matte black over-ear headphones. He puts them on, his eyebrows shoot up, and he says to camera, "okay, the bass on these is unreal." He takes them off and taps the ear cup. Natural daylight from a window. Sound: cardboard sliding, his voice, faint music leaking from the headphones.
Walking vlog
Mangosteens in a Bangkok market
Prompt
Vertical walking vlog in one take: a travel creator in a straw hat walks through a busy fruit market in Bangkok holding the camera at arm's length. She points at a pile of mangosteens and says, "you have to try these," buys one, cracks it open and takes a bite, then laughs at the camera. Bright midday light, vendors and shoppers around her. Sound: market chatter, her voice, a scooter passing.
Coaching to camera
Goblet squat cue
Prompt
Vertical shot in a sunny outdoor gym: a fitness coach in her thirties demonstrates a goblet squat with a kettlebell, then looks into the camera and says, "chest up, knees out, and breathe." She does two more slow reps. Natural sweat sheen, realistic muscle movement. Sound: her voice, the kettlebell set down on rubber flooring, birds.
Fashion try-on
Denim jacket in the mirror
Prompt
Vertical shot in a bright bedroom: a woman in her twenties stands in front of a full-length mirror trying on a cropped denim jacket over a white slip dress. She spins once, checks the back over her shoulder, then grins at her reflection and gives a thumbs up. Morning light, clothes on the bed behind her. Sound: fabric rustle, her soft laugh.
Singing
Subway busker
Prompt
Vertical shot in a subway station: a young street musician with braids plays an acoustic guitar and sings a soft, slow melody, eyes half closed. A commuter drops a coin in the open guitar case and she nods with a smile without stopping. Warm tiled walls, a train arriving in the background. Sound: her voice and guitar, the rumble of the train.
Two-person dialogue
Video podcast, two hosts
Prompt
Two hosts record a video podcast in a cozy studio with a neon sign and acoustic panels. The woman on the left leans into her microphone and asks, "so what made you quit?" The man on the right laughs, rubs his neck and answers, "honestly, I just got bored of being comfortable." Medium two-shot, warm lighting. Sound: their voices, a chair creak, no music.
Which Wan to pick
Start here
Alibaba Wan 3.0
Most clips, especially long ones
Text, a start frame or references in; up to 30 seconds with sound out. Start at 720p.
5 s clip 91 credits · 2–30 s
Open with Alibaba Wan 3.0Option 2
Alibaba Wan 3.0 Prime
Faster turnaround
The same model and options, generated faster for a higher price per second.
5 s clip 127 credits · 2–30 s
Open with Alibaba Wan 3.0 PrimeOption 3
Alibaba Wan 2.7
Matching earlier Wan 2.7 clips
The previous version, at fixed 5 or 10 second lengths. It costs more per clip than Wan 3.0, so use it only to match work you already made with it.
5 s clip 136 credits · 5–10 s
Open with Alibaba Wan 2.7Wan 2.2, the last open-weights Wan, is also in the studio's model menu at a low flat price. Credits are read from the live catalog; the Standard plan ($24/month) includes 2,000 credits, about 21 five-second Alibaba Wan 3.0 clips.
Prompting Wan well
- Always write a sound line (voices, ambience, "no music"); otherwise Wan picks a soundtrack for you.
- For long clips, write timecoded beats and keep each beat to one action.
- Be concrete: "tosses the pasta with butter and lemon zest" beats "cinematic cooking shot".
- Keep spoken lines short and in quotes, one per beat.
- Match the clip length to what happens in it: a 30-second clip with one action drifts.
- Draft at 480p or 720p, then re-run the keeper at 1080p.
From prompt to clip
Describe the scene or add a photo
Write the scene in the box at the top of this page, or start from an example. Add a photo to use it as the first frame.
Sign in and land in Video Studio
Your prompt, photo, format and the Wan version you picked carry over. Nothing runs until you press Generate.
Pick the length, then generate
Any length from 2 to 30 seconds. The credit price shows under the Generate button before anything is spent.
Check the end of long takes
Watch the last seconds of a long clip closely; re-run with a shorter length if faces or hands drift.
Questions about Wan
What is Wan 3.0 best at?
Long single takes with sound: tutorials, reviews, vlogs and short scenes that play out for up to 30 seconds without cuts.
Is Wan open source?
Wan 2.2 is, under Apache 2.0. Wan 2.5 and later, including Wan 3.0, are only available through APIs like the one Affogato uses.
What is the difference between Wan 3.0 and Wan 3.0 Prime?
Prime runs the same model with the same options, faster, and costs more per second. Use it when turnaround matters.
Does Wan generate sound?
Yes. Wan 3.0 generates dialogue, effects and ambience with the picture, on by default.
Is Wan free on Affogato?
No. Every run costs credits, priced by the second and shown before you generate. Credits come with a monthly plan and unused credits roll over up to the plan's cap.
Can I use Wan videos commercially?
Yes. Paid Affogato plans include a commercial license and no watermark, within our terms of service.
Keep exploring
Facts checked November 15, 2026. Sources: fal: how to use Wan 3.0 · fal: Wan 3.0 API · DataNorth: Wan 3.0 launch · Let's Data Science: Wan 3.0 · Alibaba Cloud: Wan 2.6 · Wan 2.2 open-source release