Wan 3.0 30-second drafts at 480p, then the 1080p final: cost on Sume
Alibaba lists Wan 3.0 at $0.05, $0.10 and $0.20 a second. A 30-second shot is $1.50, $3.00 or $6.00 at list; Sume's list x 1.25 makes it $1.875, $3.75 or $7.50.

Wan 3.0 can make a 30-second clip in one call, and Alibaba's per-second list ($0.05 at 480p, $0.10 at 720p, $0.20 at 1080p) makes a draft-then-final workflow cheap: a 30-second 480p draft is $1.50 at list and a 30-second 1080p finish is $6.00. On Sume, which bills provider list times 1.25, the same shots reserve $1.875 and $7.50 if the list is unchanged. Four drafts and one final come to $15.00 on Sume by that arithmetic.
Prices are from Alibaba Cloud's Wan3.0 announcement, read 2026-10-01, which says the figures are before promotional discounts. The Sume multiplier and the 2 to 30 second range come from the Video Generation docs.
What does each resolution cost for 30 seconds?
The Sume column is list times 1.25, shown as arithmetic. The live value is in pricing_skus from GET /v1/videos/models.
| Resolution | List per second | 30 s at list | 30 s on Sume |
|---|---|---|---|
| 480p | $0.05 | $1.50 | $1.875 |
| 720p | $0.10 | $3.00 | $3.75 |
| 1080p | $0.20 | $6.00 | $7.50 |
How should you run the loop?
Keep the prompt and duration fixed and change only resolution. A draft tells you about motion and story, not fine detail, so judge it for those. When you approve a draft, send the final with a new Idempotency-Key. Shorter duration cuts cost linearly: a 10-second draft at 480p is a third of the figures above.
Because Sume reserves the full amount on submit and describes failed jobs as released or refunded, a failed 30-second draft should not leave a lasting charge. Check the balance after a failure the first time you run this.
- Draft at 480p and 10 to 15 seconds to test the idea.
- Lock the prompt before spending on 1080p.
- Use
callback_url(HTTPS) instead of polling for 30-second jobs. - Download and keep the file as soon as the job is
completed.
Limits
The Sume figures exclude the 5.5% agent fee that is added by default. A draft at 480p can hide problems visible at 1080p, such as text rendering; Alibaba says on-screen text accuracy and audio texture are still maturing. I did not verify how long a 30-second job takes on Sume, so plan for minutes and do not block a request on it.
Sources
Related posts
More in Use cases
- Webinar highlight reel: trim five moments, stitch, add cue captions
Cut a 60-second webinar highlight reel from a recording: five trims at $0.02, one Timeline render at $0.10, and caption cues at $0.20, about $0.40 in total.
- Webinar recording to shorts: an API pipeline with Sume
Turn a webinar recording into short clips with four calls: inspect and transcribe by sentence, pick moments, trim, caption. Limits and costs included.
- Wedding highlight video: keep the vows audio, cut the rest
Cut a 2-minute wedding highlight from guest clips: take vows audio with audio-detach, place clips over it in Timeline 1.0, and add a quiet music bed.
- YouTube AI disclosure: AI backdrop and own-voice clone examples
YouTube lists cloning your own voice and extending a backdrop as edits needing no AI disclosure. What else is on that list, and what is not.
Written by Sume