Kling 3 or Wan 3.0 for a 9-second vertical ad: $1.89 vs $1.13

For a nine-second vertical ad with sound, Kling 3 costs $1.89 on Sume at either 720p or 1080p, and Wan 3.0 costs $1.13 at 720p or $2.25 at 1080p. Both make 9:16 with native audio, so the choice comes down to price at your resolution and to whether you need reference inputs.

Kling is the cheaper way to get 1080p audio at this length. Wan is cheaper at 720p and the only one of the two that takes references and runs past 15 seconds.

Price for nine seconds

Both prices below are the provider list per second times 1.25, rounded up to whole cents. Kling has one rate per audio setting. Wan has one rate per resolution.

Where they differ besides price

The Sume catalog rows are not interchangeable. The table lists what each accepts so you do not pick a model your request cannot use.

What Alibaba says about Wan 3.0

Alibaba Cloud's Model Studio documents the Wan 3.0 video model at 480P, 720P and 1080P, with a 30-second maximum, and accepts text, image, video and audio input (read 2026-10-07). Sume exposes the 2 to 30 second range and the reference limits above through its own catalog.

A practical split

If your ad is a single product shot with a voiced line and you want 1080p, send kling-3 with aspect_ratio: "9:16". If you want to hold a character look through reference images, or you want a draft at 480p first, send wan-3.0. The request shape is the same except for the model id and the references.

Either way, call GET /v1/videos/models first. It lists the exact supported_resolutions, supported_aspect_ratios and supported_durations, and /v1/videos answers a value outside them with a 400 rather than adjusting it.

• Need 1080p with sound at nine seconds: Kling 3.
• Need references or a 2-second to 30-second range: Wan 3.0.
• Need a silent Kling take: send generate_audio: false.