Wan 3.0
Alibaba's cinematic video model with sound and the highest text-to-video arena rating in NeverNo's Studio, slower to render, reached without your name or IP.
Updated November 9, 2026
A humpback whale gliding through golden clouds above a 1950s seaside town at dusk
At a glance
| Maker | Alibaba |
|---|---|
| Strengths in the app | Cinematic, Sound |
| Arena Elo (Artificial Analysis) | 1,172 |
| Typical wait | about 144 s |
| Who can use it | Paid plans |
| How it’s paid | Credits, from the first use |
| Privacy | Anonymized: NeverNo sends the request to its maker without your name, account or IP address. The maker still reads the text, applies its own rules and may keep it for a while under its own policy. |
Price
What the app charges, read from the same table the servers bill with. 1 credit = $0.01.
| 5-second clip | 126 credits ($1.26) |
|---|---|
| 10-second clip | 252 credits ($2.52) |
What it is
Wan 3.0 is a video model from Alibaba, offered in NeverNo's Video Studio on paid plans and paid in credits. It makes clips of 5 or 10 seconds in 16:9 or vertical 9:16, at 1280 by 720 pixels, with sound, from a text description or from a starting image with an optional ending image. The Studio tags it cinematic and audio. It was first of the Artificial Analysis text-to-video arena when we added it.
On that arena, read on November 7, 2026, Wan 3.0 has an Elo of 1172 from text, the highest of every video model in our catalog, ahead of Seedance 2.5 (1155), HeyGen Video 1 (1152), MiniMax H3 Max (1145) and P-Video 2 Pro (1128). From an image its Elo is 1109, below Seedance 2.5 (1158) and HeyGen Video 1 (1123), so for animating a still it is not the first pick. The conclusion is straightforward: Wan 3.0 for a scene described in words, Seedance 2.5 for a scene that starts from a picture.
The price of that rating is time. A typical 5 second clip takes about 144 seconds to render, as measured on Runware, our video provider, versus 16 for P-Video 2 Pro, 33 for HeyGen Video 1 and about 176 for Seedance 2.5. The Studio's Auto mode accounts for this: it keeps slow models for requests set to best quality, and otherwise prefers clips under about a minute. If you want Wan 3.0 for a normal request, choose it yourself.
It is anonymized. NeverNo sends the request to Alibaba without your name, account or IP address; Alibaba still reads the prompt, keeps its own red lines, and declines requests it does not accept. The finished clip is stored in your browser, not on our servers. A sensible routine uses the whole ladder of the catalog. Explore with P-Video 2 Pro or MiniMax H3 Max, whose answers arrive in well under a minute, fix the wording, the camera and the soundscape, and only then commit to a Wan 3.0 render when you want the best the arena has to show from text. Treat the wait as part of the budget: queue the render, work on something else, and come back to review. Fewer, better-planned attempts give a better film than many hurried ones. Film students, pitch writers and game studios can use the final renders as moodboards in motion, passing around a clip that conveys tone, weather and pacing far better than a paragraph of description. Keep notes on each prompt beside its output, so that when a render is finally good you know exactly which phrase earned it. Compare shadows, fabric, water and background motion across the clip, since long renders deserve a careful second viewing before you commit to a final cut. Misty mountains, bustling harbors, lantern festivals and desert caravans reward the extra patience.
Strengths
The highest text-to-video Elo in the Studio: 1172 (Artificial Analysis, read November 7, 2026).
Cinematic look with a soundtrack produced together with the picture.
Start image and optional end image, so both ends of the shot can be fixed.
720p in 16:9 and 9:16, in 5 or 10 second clips.
Limits
About 144 seconds for a typical 5 second clip, compared with about 16 for P-Video 2 Pro; trial and error is slow.
From an image its Elo is 1109, lower than Seedance 2.5 (1158) and HeyGen Video 1 (1123).
Not tagged for legible text: signs and titles may come out wrong, unlike P-Video 2 Pro or MiniMax H3 Max.
Anonymized, not private, and video needs a paid plan, paid in credits.
When to choose it
Final renders of a scripted shot
Settle the idea with a quick model such as P-Video 2 Pro, then send the polished prompt to Wan 3.0 for the version you keep: a cinematic establishing shot with ambience, ready for a film, a trailer or a pitch.
Atmosphere reels with sound
Fog over a lake, a market at dusk, a street after rain: describe the place and the sounds you expect, and obtain a clip whose picture and ambience belong together.
Start and end frame transitions
Give a starting image and an ending image, for example a seed and the full plant, or a day view and a night view of one square, and let the model create the passage between them.
Prompt tips
Write like a shot list: subject, camera, light, sound. One line each is enough, and Wan 3.0 follows structure well.
Since each try takes over two minutes, draft with a faster model first and bring only a settled prompt to Wan 3.0.
Skip lettering in the frame, or add titles afterward; this model is not tagged for legible text.
Frequently asked questions
Is Wan 3.0 the best video model on NeverNo?
From text, it has the highest arena Elo in the Studio: 1172, read November 7, 2026. From an image, Seedance 2.5 scores higher (1158 versus 1109).
How long does a clip take?
About 144 seconds for a typical 5 second clip, measured on Runware, our video provider.
Does Wan 3.0 make sound?
Yes. The Studio tags it audio, and the soundtrack comes with the picture.
Is it private?
It is anonymized: NeverNo sends the request without your name, account or IP address, but Alibaba reads the prompt and applies its own rules. Clips are stored in your browser.
Can I use start and end images?
Yes. Wan 3.0 accepts a starting image and an optional ending image, in 16:9 or 9:16.