Kling 3.0 15-second clip: $2.10 silent or $3.15 with sound on Sume

A 15-second Kling 3 clip, the longest the model makes on Sume, bills $2.10 with audio off and $3.15 with audio on. The id is kling-3, shown in the catalog as Kling Video v3 Pro. It takes 4 to 15 seconds at 720p or 1080p, in 16:9, 9:16 or 1:1. Magic Hour's tracker (read 2026-10-06) lists Kling VIDEO 3.0 as released on 2026-02-05 with 15 seconds and native audio, which matches.

The limits and rates come from the Video Router docs and the video generation docs; the vendor line is on the tracker.

Price by length and sound

Kling's list is $0.112 a second silent and $0.168 with audio, and the resolution does not change it. Sume bills list times 1.25.

What Kling 3 takes and does not

Kling 3 takes text-to-video and image-to-video with a first frame and an optional last frame. It does not take reference images, reference videos or reference audio on this route. If you need to anchor a character with references, use a model that lists supported_input_references, such as Seedance 2.5, Wan 3.0 or MiniMax H3. For a still photo driven by a motion video, Sume has a separate Kling 3.0 motion-control surface, which is its own endpoint and not kling-3.

It also does not take 21:9 or 4:3. If your delivery format is an ultrawide, Kling is out. If it is vertical or square, Kling is in.

A first request

This submits a 5-second 1080p vertical clip with sound off, which bills about $0.70. Send "generate_audio": true for sound. Always send the flag: the docs say that when it is omitted the default is the model's audio capability, and Kling can make audio, so an omitted flag can be priced as the audio-on rate.

import os, requests

r = requests.post(
    "https://api.sume.com/v1/videos",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    json={
        "model": "kling-3",
        "prompt": "A skateboarder rolls past a mural at golden hour, tracking shot",
        "duration": 5,
        "resolution": "1080p",
        "aspect_ratio": "9:16",
        "generate_audio": False,
    },
    timeout=60,
)
print(r.status_code, r.json())

When Kling is the right pick

Kling 3 is the model to try first for a short vertical or square clip where sound is optional, because the price is flat per second and does not change between 720p and 1080p. That makes 1080p cost no more than 720p here, which is worth checking against the per-resolution rates of the other models in the catalog before you assume it holds for them.

It is a poor pick for clips over 15 seconds, for reference-driven character work, and for ultrawide. For those, see the Kling 3.0 API overview and which models take references.

A note on checking the sound

If you turn sound on, listen to the clip before you ship it. Generated audio can fit the picture well and still add an effect or a murmur that you do not want in a branded video. Because the audio line adds 50 percent to the price, a sensible routine is to draft with sound off, choose the take, and rerun with sound on only for the clips whose soundtrack you will keep. A batch of ten 15-second clips costs $21.00 silent and $31.50 with sound, so the routine saves $10.50 on drafts you throw away.

Longer than 15 seconds

Kling 3 stops at 15 seconds on Sume, so a longer shot means joining two jobs on a timeline or switching to Seedance 2.5 or Wan 3.0, which go to 30. If you join Kling clips, generate the second from the last frame of the first so the scene continues, and keep the audio setting the same on both so the soundtrack does not jump.