ElevenLabs Ads Engine localizes ads in 50+ languages; what Sume covers

ElevenLabs Ads Engine and Sume solve different halves of ad localization. Ads Engine connects to your ad accounts and translates, dubs and pushes finished ads back. Sume does not connect to ad accounts, and it does not dub video in the call described here. What it does is burn translated captions into a clean video for $0.20 per clip of up to 60 seconds, so five markets cost $1.00 of caption jobs.

The ElevenLabs post (read 2026-10-05) says Ads Engine launched on June 22, 2026 inside ElevenCreative. It takes existing creatives from Google Ads, Meta Ads and LinkedIn accounts, plus performance data from running campaigns, and you choose target markets and languages (50+). It outputs localized text, adapted images, and dubbed video using Dubbing V2, which it says reproduces the speaker's tone, emotion and pacing. It also offers an optional approval step before deployment, ad fatigue detection, and reusable templates. The post did not state a price or aspect ratios.

What a team can do with Sume today

If your ad already works with a voice-over and you need readable text in each market, the caption route is the cheap one. Keep one clean master with no text burned in. For each market, send the clean master with the translated lines as cues, each with text, start and end in seconds. When you send cues, Sume does not run speech-to-text. It burns exactly your text at exactly those times, which is the right behavior for translated lines, because a speech-to-text pass would read the source language.

Each accepted job reserves and captures $0.20 for a clip of up to 60 seconds, so the cost is the number of markets times $0.20, and does not depend on the style.

A loop over the markets

The script below sends one caption job per market, with a separate Idempotency-Key for each language so a retry cannot double charge.

import os, requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
MASTER = "https://media.sume.com/artifacts/example/clean.mp4"
LINES = {
    "es": "Hecho para tu casa.",
    "fr": "Fait pour votre maison.",
    "de": "Gemacht fuer Ihr Zuhause.",
}
for lang, text in LINES.items():
    body = {"video_url": MASTER, "style": "slam",
            "cues": [{"text": text, "start": 0.5, "end": 3.5}]}
    h = {**H, "Idempotency-Key": f"ad-caption-{lang}-001"}
    r = requests.post("https://api.sume.com/v1/video-captions",
                      json=body, headers=h, timeout=30)
    print(lang, r.status_code, r.json())

Where Ads Engine is the better fit

If your problem is the loop around the ad (pulling from the platform, approving, pushing back, flagging fatigue), the ElevenLabs product describes that workflow and Sume does not. If your problem is making a text layer in many languages for a clip you already have, the caption job is the smaller tool. Latin-script languages work on the default style. The caption docs state that Latin display styles have no Hangul glyphs, and they document Hangul styles for Korean speech. They do not list faces for other scripts, so test a sample before you promise a market.

Whichever route you choose, keep the clean master. A burned-in caption cannot be removed, and the clean file is the only thing that lets you re-run one market without touching the others.

Budget the loop before you run it. Five markets is five jobs and $1.00. Ten markets is $2.00, and a clip longer than 60 seconds falls outside the priced estimate, so check the live price in the catalog first.