Allergen line on a holiday menu video: burn it in with caption cues

To put an allergen line on a holiday menu video, send the finished silent clip to POST /v1/video-captions with cues instead of words or speech-to-text. A cue is text, start and end in seconds, and Sume burns that text exactly as you wrote it at those times. Because no transcription runs, a clip with no voice does not fail, and your wording is never reworded. The part Sume does not do is check the content: the allergen statement is whatever your kitchen wrote.

This matters in the holiday weeks because a seasonal menu changes often, and the allergen line changes with it. The same silent product clip can be re-captioned for each menu without regenerating any video.

What the endpoint takes

The captions page lists the inputs below. The video must be reachable at a public video_url. You can send only one of script_text, words, cues and segments in a request, so a menu clip uses cues alone.

Steps for a menu clip

Keep the allergen line short enough to read in the time it is on screen, because the cue only controls timing and not layout judgment.

• Render or collect the silent menu clip and host it at a public HTTPS URL.
• Write one cue for the dish name and price, and a second cue for the allergen line, each with its own start and end.
• Choose a Latin style such as black-outline, then POST with a fresh Idempotency-Key per menu version.
• Poll the caption job and read the result URL, then check the burned frames by eye before posting.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: menu-nov-week-1-caption" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/example/menu-clip.mp4",
    "style": "black-outline",
    "cues": [
      {"text": "Roast squash soup, 9", "start": 0.5, "end": 3.5},
      {"text": "Contains: milk, tree nuts", "start": 3.5, "end": 7.0}
    ]
  }'

What Sume does not do

Sume does not verify allergens, prices or any claim in your cue text. It burns the characters you send. If the menu changes, send new cues; a restyle with source_caption_id reuses the earlier source video but is still a render at the same price. Check the final frames yourself, because an allergen line that is wrong is a safety problem and not a styling problem.

If your clip does carry speech and you only want the look changed, use a restyle instead of cues, and read the script text and brand spelling post for correcting words. Cues are best when the text is not what anyone says.

A practical routine for a restaurant with a rotating holiday menu is to keep one clean silent clip per dish, store the cue list for each menu version in your own notes, and re-run the caption call when a price or an allergen line changes. Because the call is idempotent per key, name the key after the menu version, such as the week and the dish, so that a double click in your tooling does not pay for the same overlay twice.