Kling 3.0 allows 3 seconds at Kling; Sume's kling-3 starts at 4

Kling's guide lists 3 to 15 seconds for Kling 3.0. Sume's kling-3 takes 4 to 15, so a 3-second beat means a 4-second clip, then a trim.

4 min readSume
All posts

If you plan a Short around Kling 3.0 and want a 3-second beat, the two ranges differ. Kling's own model guide lists a flexible duration of 3 to 15 seconds, read on 2026-10-04. Sume's kling-3 row in the Video Router catalog lists duration_seconds from 4 to 15, so a 3-second request does not match the catalog row.

The catalog is the contract, not the vendor page. Sume exposes it at GET /v1/videos/models, and a request is checked against the supported_durations of the model you name, per the video generation docs. The practical answer is to ask for 4 seconds and cut the extra second.

The 4-second clip and a trim

Submit kling-3 with duration: 4, then cut the range you want with Video trim. Trim takes a video_url on media.sume.com, a start, and exactly one of end or duration, and it returns a new MP4. Import the finished clip first, because the API does not fetch from the open internet.

import os
import requests

headers = {
    "Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
    "Idempotency-Key": "kling-beat-001",
}
body = {
    "model": "kling-3",
    "prompt": "A hand sets a ceramic mug on a wooden desk, morning light",
    "duration": 4,
    "resolution": "1080p",
    "aspect_ratio": "9:16",
    "generate_audio": False,
}
resp = requests.post("https://api.sume.com/v1/videos", headers=headers, json=body)
print(resp.status_code, resp.json().get("id"))

What the extra second costs

The registry lists Kling v3 Pro at $0.112 per second with audio off and $0.168 with audio on, and Sume reserves provider list times 1.25. That is $0.14 and $0.21 per second, so a 4-second silent clip is $0.56 and the 3-second version you keep still costs $0.56, because you pay for the generated seconds. The trim is a separate job, so budget it too. If one second of waste matters across hundreds of beats, the cheaper route is a model whose floor is lower: wan-3.0 accepts 2 to 30 seconds.

Quick comparison

Use the Kling page to decide whether the model suits the shot, and the Sume catalog to decide what you can send.

Duration floors for one clip, read 2026-10-04 (vendor page) and from the Sume registry
SourceKling 3.0 rangeFloor you can request
Kling model guide3 to 15 s3 s
Sume kling-34 to 15 s4 s
Sume wan-3.02 to 30 s2 s

Sources

Related posts

More in Models

All Models posts

Written by Sume