Wan 3.0 audio on Sume: generate_audio and audio references

Wan 3.0 on Sume can make its own audio with generate_audio or take up to 5 audio references. What to send, and how the 30-second price changes (it does not).

4 min readSume
All posts

Wan 3.0 on Sume has an optional generate_audio flag and accepts up to 5 audio references. Its Sume price is per output second by resolution; compare usage.cost on a first job with audio on and off.

The docs state that the Seedance 2.x models, Wan 3.0 and MiniMax H3 models accept audio references; Omni Flash does not.

Example

Post it to https://api.sume.com/v1/videos.

{"model": "wan-3.0",
 "prompt": "Two friends laugh at a rooftop cafe, city hum, light music",
 "duration": 15,
 "resolution": "720p",
 "aspect_ratio": "16:9",
 "generate_audio": true}

What a 15-second take costs

Wan 3.0, 15 s on Sume (read 2026-10-07)
ResolutionPrice
480p$0.94
720p$1.88
1080p$3.75

Why audio references

An audio reference lets you steer the sound of the output with an existing clip: a voice, a music bed or a room tone. Wan 3.0 on Sume takes up to 5 per job.

Use them to keep a series sonically consistent, for instance one theme across several 15-second episodes.

Costs and limits

Wan 3.0's list rate is per output second by resolution on fal ($0.05, $0.10, $0.20), and Sume bills times 1.25. The catalog does not describe a separate audio fee. Verify on your first job.

The shortest allowed Wan clip is 2 seconds and the longest 30.

Sources

Related posts

More in Models

All Models posts

Written by Sume