Add a "save this" end card to a Short with caption cues

YouTube Shorts added a Save button in July 2026, according to Wikipedia's page on the feature, which is a third-party source (read 2026-10-05). If you want viewers to use it, say so on screen. With video captions you can burn a short line into the last three seconds using cues: each cue is text, start and end in seconds, and no speech-to-text runs.

Find the end time first

A cue past the end of the clip is useless, so read the duration first. Run video inspect with "frames": false; it returns only the probe, with no stills, which includes the duration. For a 42.5 second Short, a cue from 39.5 to 42.5 is three seconds. Keep the line short enough to read in that time.

The cue request

cues and segments are phrase-level overlay text. You may send only one of script_text, words, cues and segments per request. Because you authored the text, the job also works on a silent clip, where speech-based captions fail with caption_no_speech. The fixed estimate is $0.20 for clips of at most 60 seconds (live price in GET /v1/catalog).

video_url must be a public HTTPS URL that Sume can fetch; localhost, private-network and signed URLs are refused.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: save-end-card-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/short.mp4",
    "style": "slam",
    "cues": [{"text": "Save this for later", "start": 39.5, "end": 42.5}]
  }'

What I am not claiming

I did not read a YouTube page that describes the Save button, so I will not say where it sits or how it affects reach. Sume only supplies the burned-in text; whether viewers tap Save is up to the video. Put the line clear of the platform's own buttons by setting design.placement.anchor_ratio and checking a still, as in the safe-zone post linked below.

Timing for common lengths

The cue must fit inside the clip, and the clip's real duration comes from the probe. These windows keep a three-second ask at the end. Longer lines need longer windows, so shorten the words before you lengthen the time.

Writing the line

Short beats clever. Four or five words read in three seconds: "Save this for later", "Save for your next edit". Avoid claims about what the Save button does, because the details are the platform's to describe and may change. Put the cue in the last three seconds only if the video's payoff has already landed; an ask before the payoff trains viewers to leave early.

You can add a second cue earlier, such as a one-word label on a key step, in the same cues array. Each cue needs text, start and end, in seconds, and cues should not overlap. Because no speech-to-text runs, these cues work on a silent clip, and the price is the same single caption job.

If you are unsure which duration to use, ask the probe rather than the editor: a clip exported at 42.5 seconds in your timeline can be 42.52 after the encoder rounds the last frame. Set the cue end a few hundredths early, for example 42.4, so the text is on screen to the last visible frame and the job is not asked to draw past the end of the clip.

• One request can hold several cues, each with its own start and end.
• Burned-in text is permanent in that file; keep the clean source for edits.
• Check one still with video frames before posting.