Recipe step clips: six 5 s Kling 3 shots, $4.20 silent, $6.30 sound

Six 5-second Kling 3 clips for a recipe page cost $0.70 each silent, $4.20 for six, or $1.05 each with sound, $6.30 for six, on Sume. Go silent for the steps and add one sizzle or kitchen-ambience bed in your editor, unless a particular step, such as a pour or a crack, depends on its own sound.

Why Kling 3 fits step clips

Each recipe step has a photo you already took: the chopped onion, the pan, the plated dish. Kling 3 on Sume takes that photo as the first frame and animates it for 4 to 15 seconds, at 720p or 1080p for the same price. The model has no reference inputs on Sume, so one still per clip is the unit of work.

fal's Kling v3 Pro page, read 2026-10-08, lists $0.112 per second with audio off and $0.168 with audio on. Sume's price is that list times 1.25, rounded up to the cent per job.

The numbers

Sume bills the provider list price times 1.25, rounded up to the cent per job.

Choosing where sound pays

The extra for sound is $0.07 per second. On a 5-second clip that is $0.35.

• Buy sound for the pour, the crack and the sizzle, where the audio is the proof of the step.
• Skip it for chopping and plating, where a music bed will cover the cut.
• Use the 4-second minimum for simple steps: $0.56 silent.
• Keep the same first-frame crop for all six photos, so the page reads as one set.

Submit it and read the result

Sume reserves the billable amount from your workspace USD balance when the job is submitted, then returns a polling_url. Add a frame_images entry with your step photo as the first frame, then loop over the six steps. Poll the job until status is completed or failed; a completed job lists its files in unsigned_urls.

import os
import time
import requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
body = {
    "model": "kling-3",
    "prompt": "A slow push-in on the scene, soft natural light",
    "duration": 5,
    "resolution": "720p",
    "aspect_ratio": "16:9",
}
r = requests.post("https://api.sume.com/v1/videos", headers=H, json=body, timeout=60)
r.raise_for_status()
url = r.json()["polling_url"]
while True:
    job = requests.get(url, headers=H, timeout=30).json()
    if job["status"] in ("completed", "failed"):
        break
    time.sleep(10)
print(job["status"], job.get("unsigned_urls"))

What to check

Food video is forgiving of texture and unforgiving of hands. Check fingers and knife handling in a silent 4-second test first. I did not test how the model renders hot steam or flames and have no vendor statement about them; state quantities, temperatures and times in the recipe text, not in the video.