At 4 credits a second at 480p, a 30-second Wan 3.0 draft costs 120 credits.
How to use Wan 3.0 on Nsketch: 2 to 30 second clips from 4 credits a second, reference limits, a prompt pattern, a worked example and Prime.
What Wan 3.0 Is Good At
Wan 3.0 is Alibaba's video model for combining image, video and audio references into one clip. On Nsketch it renders 2 to 30 seconds at 480p, 720p or 1080p, with sound.
Price is the main reason to choose it. At 4 credits a second at 480p, a 30-second draft costs 120 credits, so you can run a long scene several times before committing to a final. It suits trend and dance clips that put a person from a photo into the moves of a reference video, long travel shots, and clips whose sound should follow music you upload.
Watch the reference limits. Reference videos and audio clips are each capped at 15 seconds in total, and the output plus any reference video can't pass 30 seconds, so a 15-second reference leaves room for 15 seconds of output.
Wan 3.0 Settings on Nsketch
Set the length to auto when you copy a dance or a camera move, so the output runs exactly as long as the reference clip.
| Setting | Wan 3.0 on Nsketch |
|---|---|
| Length | 2 to 30 seconds, or auto to match your reference video |
| Resolution | 480p, 720p or 1080p |
| Aspect ratio | 16:9, 9:16, 1:1, 4:3 or 3:4 |
| Audio | On or off, same price |
| Reference images | Up to 10 |
| Reference videos | Up to 5, 15 seconds in total |
| Audio references | Up to 5, 15 seconds in total |
| First and last frame | Supported, instead of reference media |
| Versions | Wan 3.0 and the higher-detail Wan 3.0 Prime |
How to Prompt Wan 3.0
With references, say what each one supplies: the person from the reference photo, the dance from the reference video, the music from the reference audio. Then describe only what is new, such as the location, the light, and the framing if it should differ from the reference.
Without references, describe one subject, one continuous camera move with an ending, and the sounds you want. Wan 3.0 generates audio, so a short list of sources like "soft acoustic guitar, wind, distant trams" shapes the track more than a mood word.
- Subject: who or what, and the one action they perform
- Place: two or three concrete details, such as terracotta roofs and a river
- Camera: one move and where it ends
- Sound: two to four named sources
- Look: light or style in a few words
- References: what each upload supplies, if you added any
Worked Example: A Rooftop Dance in Lisbon
The prompt is four sentences. The first sets the subject and place with concrete details: a white dress, terracotta roofs, the river. The second gives the camera one continuous move with a finish, circling her and then rising to reveal the city.
The last two sentences carry sound and light. Three named sources give the track a sense of place, and "warm cinematic light" is the only style note. No references were needed for a text-only shot like this.
Wan 3.0 · 8 seconds · 1080p · 16:9 · with sound
Prompt
A dancer in a flowing white dress spins across a sunlit rooftop in Lisbon, terracotta roofs and the river behind her. The camera circles her in one continuous take, then rises to reveal the city. Soft acoustic guitar, wind, distant trams. Warm cinematic light.
What Wan 3.0 Costs in Credits
Wan 3.0 bills per second: 4 credits at 480p, 8 at 720p and 14 at 1080p, with sound on or off at the same price. The example above, 8 seconds at 1080p, costs 8 × 14 = 112 credits.
Reference video seconds are billed along with the output. A 10-second clip at 720p guided by a 10-second reference video bills 20 seconds, or 160 credits. The generator shows the total before you generate.
| Version | 480p | 720p | 1080p |
|---|---|---|---|
| Wan 3.0 | 4 credits/s | 8 credits/s | 14 credits/s |
| Wan 3.0 Prime | 6 credits/s | 12 credits/s | 26 credits/s |
When to Pick a Different Model
Pick Wan 3.0 Prime, in the same model picker, for final renders that need more detail. It takes the same inputs, and the example clip at 1080p would cost 208 credits instead of 112.
Pick Seedance 2.5 when your references outgrow Wan's caps. It accepts up to 30 images, 3 videos and 10 audio clips, allows 30 seconds of reference video or audio in total, and adds 21:9.
Pick Veo 3.1 for an 8-second clip built around a written line of dialogue, or Seedance 2.0 Standard when the delivery file has to be 4K.
Specs, sample prompts and pricing notes are on the Wan 3.0 model page.
FAQ
Frequently asked questions
It makes the output as long as your reference video, which is what you want when copying a dance or a camera move.
The output and the reference video share a 30-second limit. A 15-second reference leaves room for up to 15 seconds of output.
Find the prompt and timing with Wan 3.0, then switch to Prime for the version you publish. Prime takes the same inputs and costs 6 to 26 credits a second.
No. A generation uses either first and last frames or reference media, not both.
No. Audio on or off costs the same per second.