Wan 3.0
Wan 3.0 is Alibaba Tongyi Lab's newest video model, in public beta on Alibaba Cloud's Model Studio since 6 August 2026 and on fal since late August. It doubles the family's ceiling to 30 seconds in a single pass, keeps native audio, adds documents and web pages as references alongside up to 20 images, clips and audio tracks, and comes in two tiers on fal: standard at $0.05, $0.10 and $0.20 a second by resolution, and Prime, about 40 percent more, for faster turnaround. Like 2.7 it is hosted only; there are no weights.
Anvisha Pai, Co-founder & CEO, Voyager
Alibaba · released
›What we asked for
A self-portrait of Wan 3.0, Alibaba's video model, as a single continuous shot: a paper lantern lifts from a night market in Hangzhou, the camera rising with it over the canal until the whole city of lights is below, a soft erhu line and crowd murmur under it.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 5, audio: true
returned 1280 by 720 pixels
At a glance
- Price
- $0.10 a second at 720p, $0.05 at 480p, $0.20 at 1080p; Prime is $0.14 at 720p. Our five-second clips cost $0.50.
- Length
- 2 to 30 seconds in one pass, or leave duration unset and let the model choose.
- Audio
- Native, on by default, switchable off.
- Sizes
- 480p, 720p or 1080p in five shapes plus adaptive, which follows a starting image.
- Thinking
- An optional reasoning pass over composition and motion before rendering.
Other Wan models: Wan 2.7. Compare all 2.
Anvisha's take
Wan 3.0 is the best value on the roster for ordinary shots. Every clip is steady and consistent: the turntable holds its shape, the speaking woman is natural and lit like a real kitchen, the spill throws real water, the market is dense without melting, and the woodblock lighthouse gets a proper storm. The self-portrait, a lantern rising over a canal town until the whole city is below, is the only hero that did exactly what the prompt asked. The one failure is inherited from 2.7: fal's content checker refused the festival title card again. $0.50 a clip at 720p, about three minutes each, 30 frames a second. The September 18 workshop tests below returned both 20- and 30-second clips. The sampled frames keep the red mug and blue apron recognizable and reach the final look toward the window; audio and full-shot continuity were not formally scored.
Wan 3.0 real output: the same 9 clips every model gets
Every video model on this site gets the same 9 clips, so you can compare like with like. These are the clips Wan 3.0 returned, one try each, nothing re-rolled or retouched. Each tile shows one frame of the clip until you press play. Generated on . The 8 clips took 128 to 194 seconds each and cost $4.00 in total.
A clean product shot that holds its shape through a full slow rotation.
›What we asked for
A matte black ceramic pour-over coffee dripper rotating slowly on a white turntable, soft studio light, seamless white background, product video, no text.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 5, audio: true
returned 1280 by 720 pixels
A face, lip sync and a spoken line, where the model makes audio.
›What we asked for
A woman in her 40s with short grey hair stands in a bright kitchen, looks at the camera and says 'Good morning, the coffee is ready', natural window light, handheld feel, she smiles at the end.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 5, audio: true
returned 1280 by 720 pixels
Set text that stays spelled correctly while it moves.
›What we asked for
A title card reading 'MODA SUMMER SESSIONS' animates in over slow-motion footage of a festival crowd at dusk, then a smaller line 'Friday 12 June' fades in beneath it. Bold condensed type, tangerine on navy.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 5, audio: true
A steady drone push-in with stable geometry.
›What we asked for
Drone shot pushing slowly forward over a coastal town at sunrise, terracotta roofs, fishing boats in the harbour, long shadows, cinematic and steady.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 5, audio: true
returned 1280 by 720 pixels
A spill in slow motion: does water behave like water.
›What we asked for
A glass of water tips over on a wooden table and spills in slow motion, water and three ice cubes sliding across the wood, macro lens, morning light.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 5, audio: true
returned 1280 by 720 pixels
Hands, a kettle and a steady pour without extra fingers or teleporting objects.
›What we asked for
Close-up of hands pouring hot water from a gooseneck kettle in slow circles into a pour-over coffee dripper, steam rising, shallow depth of field, warm cafe light.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 5, audio: true
returned 1280 by 720 pixels
Many people and objects moving at once without merging.
›What we asked for
A crowded Saturday farmers market seen from slightly above, shoppers with tote bags, a dog on a leash, a busker with a guitar, bunting swaying, late morning light, gentle camera drift.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 5, audio: true
returned 1280 by 720 pixels
A named style held across every frame rather than drifting to photoreal.
›What we asked for
A lighthouse on a rocky cliff in a storm, in the style of a Japanese ukiyo-e woodblock print, waves crashing in flat colour with visible wood grain, animated like moving paper cutouts.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 5, audio: true
returned 1280 by 720 pixels
Bring a still product photo to life without changing the product.
›What we asked for
The coffee dripper slowly fills as coffee drips through it, steam rising, the camera drifting gently around it, the dripper itself unchanged. (starting from The Nano Banana Pro product shot from the image fixture set.)
alibaba/wan-3.0/image-to-video
resolution: 720p, duration: 5, audio: true
returned 960 by 960 pixels
How the cost under each picture was worked out: fal's Wan 3.0 rate of $0.10 per second at 720p (fal model page, read 2026-09-09) for the 5 seconds requested; fal does not return the charge.
Additional workflow tests
Separate from the common benchmark above. These tests examine longer shots, successive edits or specialized controls. One attempt per test, including failures. Costs are estimates from published rates.
20-second continuous workshop shot
Wan 3.0
One unbroken shot, stable apron, mug and window; the action must reach the final pose.
›What we asked for
One continuous 20-second take in a sunlit pottery workshop, no cuts. A woman wearing a plain blue apron stands at a wooden table with one red mug and a folded white cloth. First third: she picks up the red mug and turns it once in her hands. Middle third: she places it back on the table and wipes beside it with the white cloth. Final third: she folds the cloth, places it to the left of the mug, and looks toward the same window. The camera slowly tracks sideways throughout. Keep the same woman, apron, red mug, table and window. Natural workshop ambience, no music, no text.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 20, audio: true
returned 1280 by 720 pixels
Generated . Cost basis: fal's Wan 3.0 rate of $0.10 per second at 720p (fal model page, read 2026-09-09) for the 20 seconds requested; fal does not return the charge; rate also verified 2026-09-18 at https://fal.ai/models/alibaba/wan-3.0/text-to-video. Original output.
Returned 20.04 seconds at 1280x720, estimated $2.00, after about 458 seconds. The sampled frames retain the apron, mug and workshop and reach the window-facing pose. This is frame review, not proof of uninterrupted continuity.
30-second continuous workshop shot
Wan 3.0
One unbroken shot, stable apron, mug and window; the action must reach the final pose.
›What we asked for
One continuous 30-second take in a sunlit pottery workshop, no cuts. A woman wearing a plain blue apron stands at a wooden table with one red mug and a folded white cloth. First third: she picks up the red mug and turns it once in her hands. Middle third: she places it back on the table and wipes beside it with the white cloth. Final third: she folds the cloth, places it to the left of the mug, and looks toward the same window. The camera slowly tracks sideways throughout. Keep the same woman, apron, red mug, table and window. Natural workshop ambience, no music, no text.
alibaba/wan-3.0/text-to-video
aspect_ratio: 16:9, resolution: 720p, duration: 30, audio: true
returned 1280 by 720 pixels
Generated . Cost basis: fal's Wan 3.0 rate of $0.10 per second at 720p (fal model page, read 2026-09-09) for the 30 seconds requested; fal does not return the charge; rate also verified 2026-09-18 at https://fal.ai/models/alibaba/wan-3.0/text-to-video. Original output.
Returned 30.02 seconds at 1280x720, estimated $3.00, after about 794 seconds. The middle sampled frame shows wiping beside the mug and the late frame shows the cloth to its left. The camera framing changes substantially during the shot.
Compare Wan 3.0
Wan 3.0 pricing
| Where | Unit | Price | Source |
|---|---|---|---|
| fal | Second at 720p $0.05 at 480p, $0.20 at 1080p, the same on the text, image and reference endpoints. Prime: $0.068, $0.14 and $0.28. | $0.10 | fal model page, read |
| Alibaba Model Studio | Second at 720p List price; a promotion until 24 September 2026 charges $0.035, $0.07 and $0.14 a second at 480p, 720p and 1080p. | $0.10 | The Rundown on Wan 3.0 pricing, read |
| fal | Five-second 720p clip | $0.50 | fal model page, read |
Is Wan 3.0 free?
Alibaba's Wan app gives daily free credits, and fal's Wan 3 landing page advertises five free 720p text-to-video generations a day for signed-in users. The API charges per second, and there are no weights to run yourself.
- •Wan app: daily free credits.
- •fal: five free 720p text-to-video clips a day on the Wan 3 page for signed-in accounts, read 9 September 2026.
- •Alibaba Model Studio and hosts: per second, promotional rates until 24 September 2026.
- •Open weights: none; the newest downloadable Wan video weights are 2.2.
fal's Wan 3 page, read
Where to use Wan 3.0
- Wan app · Alibaba's consumer app with daily credits.
- Alibaba Model Studio · Alibaba Cloud's hosted API, promotional pricing until 24 September 2026.
- fal · Per second by resolution; text, image and reference endpoints, plus a Prime tier.
- Voyager · Runs inside Voyager, the agent for creative work. Private preview by waitlist.
Voyager
Run Wan 3.0 inside Voyager
Voyager is an agent for creative work that runs every model in this reference, Wan 3.0 included. Brief it, and it picks the model, makes the thing and hands back something you can edit. Create an account to get started.
What we noticed running Wan 3.0
- •Refused, as on Wan 2.7: fal's content checker returned a 422 on the festival title card prompt. No other vendor's checker objected to it. See the animated title card clip.
- •Followed the square starting image and came back 960 by 960 with aspect_ratio left adaptive; the pour lands cleanly in the dripper. See the image to video clip.
- •The glass tips, spills and rolls with the ice cubes staying cube-sized, fixing the scale drift 2.7 showed on the same prompt. See the physics clip.
- •Self-portrait: the lantern rises from the crowd to a city-wide aerial in one continuous move, the closest any model came to its own hero prompt.
- •Clips come back at 30 frames a second where most of the roster returns 24; waits were 128 to 194 seconds.
Wan 3.0 limits and API parameters
| Limit | Value |
|---|---|
| Length | 2 to 30 seconds in one pass. fal API reference, read |
| Resolution | 480p, 720p or 1080p. fal API reference, read |
| References | Up to 10 images, 5 clips and 5 audio tracks on fal's reference endpoint; Alibaba's own API takes up to 20 plus documents and web links. fal's Wan 3 page, read |
›Every setting, for developers
fal (alibaba/wan-3.0/text-to-video)
fal API reference, read
| Parameter | Type | Values | Default |
|---|---|---|---|
| prompt | string | Any text Required. | |
| duration | integer | 2 to 30 seconds, or unset to let the model choose | 5 |
| resolution | enum | 480p, 720p, 1080p | 1080p |
| aspect_ratio | enum | adaptive, 16:9, 4:3, 1:1, 3:4, 9:16 | adaptive |
| audio | boolean | true, false | true |
| enable_thinking | boolean | true, false Reasons about composition and motion before rendering. | |
| enable_prompt_expansion | boolean | true, false | true |
| seed | integer | Any integer | |
| enable_safety_checker | boolean | true, false | true |
| start_image_url | URL | Starting frame Image-to-video endpoint; end_image_url sets a closing frame. |
What Wan 3.0 will not make
Wan's hosted models run behind fal's safety checker, on by default and only switchable off with account authorisation, and under Alibaba Cloud's Model Studio terms. The practical boundaries are sexual content, minors, violence and deceptive depictions of real people. In our run the checker refused a festival title card on the 2.7 endpoint.
fal API reference, read
Alternatives to Wan 3.0
Frequently asked questions
What is Wan 3.0?
Wan 3.0 is Alibaba Tongyi Lab's video model, in public beta since 6 August 2026, making clips of 2 to 30 seconds in one pass with native audio, at 480p to 1080p. It is on Alibaba Model Studio and fal, in standard and Prime tiers.
How much does Wan 3.0 cost?
$0.05, $0.10 and $0.20 a second at 480p, 720p and 1080p on fal, with Prime at $0.068, $0.14 and $0.28, read on 9 September 2026. Alibaba's own API runs a promotion at $0.035, $0.07 and $0.14 until 24 September 2026.
Is Wan 3.0 open source?
No. It is an API-only beta with no published weights or licence; the newest downloadable Wan video weights are 2.2.
Wan 3.0 or Wan 2.7?
3.0 for anything over 15 seconds or that needs references; the per-second price is the same. Both pages run the same nine five-second clips, so compare the tiles before deciding.
How this page is made
Prices, limits and settings are read from the linked vendor and host pages on the dates shown. Every picture was generated by us through fal with the request shown under it, and the original files are kept unedited. A generation that failed is shown as a failure. Model and vendor names are used to identify the products; no vendor imagery appears on this page.
