Wan 3.0 reference clips: four 4-second videos break the 15 s cap

Wan 3.0 on Sume accepts up to 5 reference videos, with a combined length of at most 15 seconds and a frame rate of at least 16 fps. Four 4-second clips add up to 16 seconds and break the cap; three 5-second clips add up to exactly 15 and pass. Count the total before you submit, not the number of clips.

Which combinations pass

The rule has two parts: at most 5 clips, and at most 15 seconds in total. A single 15-second clip is allowed.

Related caps on the same row

Reference images are capped at 10 and reference audio at 5 clips with 15 seconds in total. The output length is separate: 2 to 30 seconds. At 720p a 10-second output costs 10 x $0.125 = $1.25, and the catalog constraints list no extra fee for references on Wan.

• Check the clip's frame rate with video inspect before you spend money.
• A phone clip at 24 or 30 fps passes the 16 fps floor; screen recordings at a low variable rate may not.
• Trim a clip to a safe length with video trim, $0.02 per job.

A preflight check

The check below fails fast. It takes durations you already know, for example from video inspect.

def check_wan_refs(clips):
    # clips: list of (seconds, fps)
    if len(clips) > 5:
        return "more than 5 reference videos"
    if sum(s for s, _ in clips) > 15:
        return "reference videos exceed 15 seconds in total"
    if any(fps < 16 for _, fps in clips):
        return "a reference video is below 16 fps"
    return "ok"

print(check_wan_refs([(4, 24)] * 4))
print(check_wan_refs([(5, 24)] * 3))

What the error looks like

A request that breaks a reference limit is refused before generation, so the usual cost is time rather than money; failed jobs are refunded. Read the error field on the poll response, and make sure every reference is available over public HTTPS.

Keep a manifest of your reference clips with duration and fps, and re-run the check whenever you swap one. A 15-second budget goes quickly: one long establishing shot and two short action shots is a better split than five tiny clips.