Pressure-wash and gutter reminder: one clip, six neighborhood captions

Burn a different neighborhood name into one silent clip with Sume video captions cues, $0.20 per version. Six versions cost $1.20 on top of the base clip.

4 min readSume
All posts

To send a fall gutter or pressure-wash reminder to six neighborhoods, make one silent base clip and burn a different line of text into each copy with Sume video captions. Each version is $0.20, so six come to $1.20 on top of the base clip.

Neighbors respond to their own street. A clip that says "Maple Ridge, we are in your area Oct 21-23" outperforms a generic one, and it costs only a caption job to change. This works because captions can carry authored text without any speech.

The cues path for silent clips

Normal captions transcribe speech. A silent clip fails with caption_no_speech. The fix in the video captions docs is to send cues (or segments), each with text, start and end in seconds. Sume burns the text at those times and does not run speech-to-text. You can send only one of script_text, words, cues and segments.

The input is a video_url that is a public HTTPS link, and the clip can be up to 60 seconds for the $0.20 price.

import os, requests

BASE = "https://api.sume.com"
HEAD = {"Authorization": "Bearer " + os.environ["SUME_API_KEY"]}
clip = "https://example.com/gutter-base.mp4"  # public HTTPS URL

for name in ["Maple Ridge", "Cedar Park", "Old Mill"]:
    body = {
        "video_url": clip,
        "cues": [
            {"text": "Gutter cleaning, " + name, "start": 0.5, "end": 3.0},
            {"text": "Oct 21-23, text us to book", "start": 3.0, "end": 6.0},
        ],
    }
    r = requests.post(BASE + "/v1/video-captions", json=body, headers={**HEAD, "Idempotency-Key": "gutter-" + name}, timeout=60)
    print(name, r.status_code)

Making the base clip

The base clip can be a 6-second Wan 3.0 clip at 720p ($0.75) from a photo of a clean roofline, or footage you already have, provided it is reachable at a public HTTPS URL. Keep the movement slow and the lower third empty so the cues have space.

Sume's Wan 3.0 pricing is $0.125 per second at 720p (read 2026-10-10, Video Generation docs). Add the six caption jobs and the total is $0.75 + 6 x $0.20 = $1.95.

Six-neighborhood reminder set (read 2026-10-10)
ItemCountUnitSubtotal
Base clip, Wan 3.0 720p, 6 s1$0.75$0.75
Caption job with cues6$0.20$1.20
Total$1.95

Check the destination first

Google Business Profile accepts videos up to 30 seconds, up to 75 MB, 720p or higher, per Google's help page (read 2026-10-10, source). A 6-second clip clears all three, so the same file can go to the profile and to a social post.

If you plan dozens of neighborhoods, use the bulk path instead of a loop. The earlier post on a bulk run ad matrix shows how to name items so you can map results back.

Writing cues that fit the frame

Cue text has to be short. A reader gets about two seconds for a line on a phone, so one idea per cue: the service and the place first, the dates and the way to book second. Avoid cues that stretch past the width of a vertical frame, and leave the first half second empty so the first frame is clean when the clip loops.

Choose a caption style that is readable on a busy outdoor background. The docs list slam, punch and tiktok-green for Latin text. Test one neighborhood, look at the render, and only then fan out to the rest. If you do not like the look, a restyle of an existing caption job keeps the same text without a second transcription, but with authored cues the cheapest fix is simply a new request at the same $0.20.

A tidy habit is to keep a spreadsheet with one row per neighborhood, the cue text, the idempotency key and the returned job reference, so a retry never double-bills and a missing result is easy to chase.

  • One idea per cue, five words or fewer where you can.
  • Keep cues inside the safe area, away from platform buttons.
  • Reuse the same idempotency key on a retry of the same neighborhood.

What to watch

Do not promise availability dates you do not hold. Re-read each cue against your schedule before sending. The caption job burns text exactly as written, including typos.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume