Wan 3.0 from Alibaba makes one take of up to 30 seconds and adds the sound in the same pass. Write a prompt or start from a photo.
Every idea names what you hear as well as what you see. Click one to load it into the generator, then make it yours.
IllustrationText to video · 16:9 · 5 s
Prompt: Night street in heavy rain: a violinist plays a slow melody under a lamp post, umbrellas pass behind him, the camera pushes in slowly. Sound: violin melody, steady rain on the pavement, distant footsteps
IllustrationText to video · 16:9 · 5 s
Prompt: Wide shot of a stormy coast: big waves crash against black rocks and throw spray into the air, dark clouds roll overhead, a beam of light breaks through. Sound: crashing waves, howling wind, low thunder at the end
IllustrationText to video · 16:9 · 5 s
Prompt: Handheld walk through a busy night market: steam rises from food stalls under red lanterns, a cook turns skewers on a grill, people pass by. Sound: sizzling grill, crowd chatter, a vendor calling out once
IllustrationText to video · 16:9 · 5 s
Prompt: Slow aerial shot at sunrise: a colorful hot air balloon rises over green hills and morning mist, the burner flares once. Sound: soft wind, the roar of the burner, birds waking up
Wan 3.0 is Alibaba's AI video model, released in August 2026. It turns a prompt, a photo, or a first and last frame into one continuous take of up to 30 seconds, and it makes the sound in the same pass, so rain, footsteps and engines follow what happens on screen.
On Imgveo, Wan 3.0 renders at 480p, 720p or 1080p and costs 16 credits per second at 720p. Sound is part of that price, not an add-on, and failed clips are refunded.
Wan 3.0 Prime is the faster version of the same model. It has the same lengths, resolutions and sound, and costs more per second, so it suits deadlines better than drafts.
What you can generate right now. Every number comes from the generator's own settings.
Not an illustration: we generated this clip on Imgveo with the prompt and settings below, and nothing was edited afterwards. Turn the sound on.
Prompt
A dog barks twice at a passing bicycle on a quiet street, birds chirping, a car horn in the distance
The prompt asks for two barks, birdsong and a car horn in the distance. Play it with sound and judge how closely Wan 3.0 followed.
The same family, twice the length, and sound for the first time. Both models run on Imgveo, and every row below comes from the generator's settings.
| Wan 2.7 | Wan 3.0 | |
|---|---|---|
| Longest take | 15 s | 30 s |
| Resolution | 720p / 1080p | 480p / 720p / 1080p |
| Sound | None | Made in the same pass |
| First and last frame | Up to 5 s | Up to 30 s |
| Credits per second, 720p | 16 | 16 |
One take now runs up to 30 seconds instead of 15, so a scene can build and land without a cut.
Wan 2.7 clips were silent. Wan 3.0 adds ambience and effects that follow the picture, at no extra cost.
First and last frame mode now runs to 30 seconds; on Wan 2.7 it stopped at 5.
480p at 8 credits per second lets you test an idea for half the 720p price.
All three run on Imgveo. Wan 3.0 and Prime share every setting; Prime is faster and costs more. Wan 2.7 stays for silent clips.
The default: drafts at 480p, finished clips at 720p or 1080p, up to 30 seconds with sound.
5 s at 720p: 80 credits
The faster version. Pick it when the turnaround matters more than the credits.
5 s at 720p: 130 credits
Silent clips of up to 15 seconds at 720p or 1080p, for jobs that already work well on it.
5 s at 720p: 80 credits
| Wan 3.0 | Wan 3.0 Prime | Wan 2.7 | |
|---|---|---|---|
| Longest take | 30 s | 30 s | 15 s |
| Sound | Yes | Yes | No |
| Credits per second, 480p | 8 | 13 | — |
| Credits per second, 720p | 16 | 26 | 16 |
| Credits per second, 1080p | 32 | 51 | 24 |
Credits per second of video. Render times depend on the length, the resolution and the queue.
No editing software and no setup: the generator at the top of this page does the whole job.
Pick Wan 3.0 or Prime, then write a prompt, upload one photo as the first frame, or upload a first and a last frame.
Say what the camera sees and what happens, then add a short sound line. The Sound Cue method below shows how.
Choose 5 to 30 seconds and 480p, 720p or 1080p; the button shows the exact credits. Download the MP4 with its soundtrack.
Wan 3.0 builds the soundtrack from your words. Leave sound out and it guesses; name it in four layers and the clip sounds the way you pictured it.
The shot and the place: "Night street in heavy rain, a violinist under a lamp post."
What moves and when: "He plays a slow melody while umbrellas pass behind him."
The bed of sound that never stops: "steady rain, distant traffic."
One or two short events tied to a moment: "a car door slams once near the end."
Vague
A beach at sunset, relaxing vibe
With sound cues
Wide shot of a beach at sunset, small waves roll in, a dog runs along the waterline. Sound: soft waves, a dog barking twice, gulls in the distance
"Rain on a tin roof" gives the model something to make. "Moody atmosphere" does not.
"Barks twice" or "thunder at the end" places the sound; "barking" can fill the whole clip.
A crowded sound line turns into noise. Keep one ambience and one or two events.
End the prompt with "Sound: …" so the soundtrack never gets mixed up with what should be on screen.
Shot, subject and action, setting and light, then sound. Four parts, in that order, cover almost any Wan 3.0 clip.
Shot size and camera move first: wide, close-up, handheld, aerial or locked-off.
Who or what is on screen and what they do. Plan one beat for every 5 to 10 seconds.
Where and when: a rainy street at night, a kitchen at noon, hills at sunrise.
One ambience and one or two spot sounds, after "Sound:" at the end.
Replace the words in brackets. Each template is written for the length shown.
Text to video · 16:9 · 10 s
[Close-up, slow push-in]: [dish] being [cooked or served] in [kitchen or restaurant], [steam, flames or sauce detail]. Ends on the finished plate. Sound: [cooking sound], [kitchen ambience], [one spot sound]
Text to video · 16:9 · 15 s
[Handheld walk] through [place] at [time of day], [weather], [people or traffic detail]. The camera turns toward [landmark]. Sound: [ambience], [distant sound], [one spot sound]
Image to video · 9:16 · 10 s
The person in the photo [action] in [setting]. The camera [move]. At the end they [second action]. Sound: [ambience], [a sound tied to the action]
Text to video · 16:9 · 30 s
[One continuous tracking shot]: [subject] [first beat]. Then [second beat]. Then [third beat]. Ends with [final image]. Sound: [ambience throughout], [spot sound at the second beat], [spot sound at the end]
Without a sound line Wan 3.0 picks its own soundtrack. Name at least the ambience.
Three events in a 5-second clip feel rushed. Save multi-beat stories for the longer lengths.
Wan 3.0 stops at 1080p. For a 4K file, upscale the finished clip instead of re-rendering.
You pay per second of video, and the sound costs nothing extra. Every number below comes from the same code that prices the generator.
| Length | 480p | 720p | 1080p | Prime 720p |
|---|---|---|---|---|
| Per second | 8 | 16 | 32 | 26 |
| 5 seconds | 40 | 80 | 160 | 130 |
| 10 seconds | 80 | 160 | 320 | 260 |
| 15 seconds | 120 | 240 | 480 | 390 |
| 20 seconds | 160 | 320 | 640 | 520 |
| 30 seconds | 240 | 480 | 960 | 780 |
Credits per clip. Sound is included.
Credits = seconds × the per-second rate. A 10-second 1080p clip costs 320 credits; the same clip on Prime costs 510.
The Starter plan's 1,500 monthly credits cover about 18 five-second Wan 3.0 clips at 720p, or 37 at 480p.
If a generation fails, the credits go back to your balance automatically. You only pay for the videos you receive.
For the full cost math, including retries, read AI video credits explained
Four jobs where a long take with its own soundtrack saves the most work. The pictures illustrate each idea; they are not Wan 3.0 output.
IllustrationA flaming wok, rising steam and kitchen noise sell a dish better than a silent loop.
IllustrationTurn a backyard photo into a moving scene with laughter and birdsong, ready for a reel.
IllustrationA finish line and a cheering crowd, with up to 30 seconds to build to the moment.
IllustrationA snowy cabin at dusk with the wind and the crackle of a fire, for a listing or a teaser.
Open Imgveo Studio to run this and every other AI video and image model from a single workspace, with one credit balance.
Compare Wan 3.0 with other models, or take a finished clip further.
Describe the picture and the sound, choose 5 to 30 seconds, and generate. Failed clips are refunded, and every video downloads without a watermark.
Wan is a trademark of Alibaba Group. Imgveo is an independent platform and is not affiliated with Alibaba.