Perfume ad shot at 1080p: H3 Max $2.00 vs Kling 3 $1.40 for 10 s

At 1080p, a 10-second shot costs $2.00 on MiniMax H3 Max and $1.40 on Kling 3 silent ($2.10 with sound), on Sume. Pick Kling 3 when you start from one hero still of the bottle. Pick H3 Max when you want to send several angles of the bottle, a mood video and a sound reference as inputs, because Kling 3 on Sume takes no reference inputs.

What each model takes

Perfume ads live on the object: glass, label, liquid and light. Consistency of the bottle across shots matters more than the background. Kling 3 on Sume accepts a first frame and an optional end frame, which keeps one bottle consistent for a single shot. H3 Max accepts up to 9 reference images, 3 reference videos and 3 reference audio files, with 12 in total; audio cannot be the only reference.

MiniMax's H3 post, read 2026-10-08, describes video, image and audio inputs in one model and native stereo sound. Sume's catalog notes that H3 Max's 1080p comes from a latent refinement of a 768p base, not native 1080p generation.

Per-job comparison

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

A pragmatic route

The cheapest path to a finished shot is not the cheapest model.

• Draft on H3 at 768p ($0.75) with your bottle photos as references to find a camera move you like.
• Finish on Kling 3 ($1.40) using the best H3 frame as the first frame, if you want 1080p at the lower rate.
• Or finish on H3 Max at 1080p ($2.00) if the references should keep driving the shot.
• Check the bottle label and typography frame by frame; generated text drifts.

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. Swap the model to kling-3 and add a first frame to compare the same shot. 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": "minimax-h3-max",
    "prompt": "A slow push-in on the scene, soft natural light",
    "duration": 10,
    "resolution": "1080p",
    "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 I did not compare

I did not benchmark image quality between the two. The comparison above is limited to price, resolution, length and reference inputs, all from the Sume catalog and the pages in sources. Take the rendering judgment from your own side-by-side test of the same bottle.