Captions app background stroke and templates vs Sume design

What did the Captions app add in June 2026?

The Captions Help Center changelog lists two June 2026 entries: Caption Background Stroke, added for readability, and Cloud-Backed Custom Caption Templates that sync across devices. It also lists Phrase Editing on Web (April 2026), which lets you edit captions by phrase in the web editor. The page does not describe the stroke's controls in detail, so this post only compares what the entries say.

If you build caption looks in an app and want the same look from code, the questions are: can you set an outline, can you save the look, and can you reuse it on the next clip without a second transcription? Sume answers yes to the first and the third through its API, and leaves the second to you.

How does Sume handle stroke and card backgrounds?

Standalone video captions take a design object that merges over the chosen style. Under colors there are stroke and card (set card to null to draw no card); under typography there are stroke_width_px, safe_width_ratio, font_size_ratio, base_weight, active_weight and active_scale. Placement uses anchor_ratio and landscape_anchor_ratio, phrasing uses max_words, max_chars and pause_seconds, and motion has four timing fields.

Colours are hex, rgb()/rgba() or transparent; other CSS syntax is rejected. Numbers outside their documented range return a 400 at request time, so a bad look fails before rendering or billing. One restriction matters: design is not supported on the punch and tiktok-green styles.

How do you save a caption look on Sume?

There is no server-side template library in the docs for captions. A look is a request body, so the practical equivalent of a cloud template is a JSON file in your repository or your automation tool, merged into each request. That is arguably better for teams: it is versioned, reviewable and the same across every machine.

This body keeps black-outline, adds a six-pixel stroke, a semi-transparent card, a yellow spoken-word colour and short four-word phrases.

{
  "video_url": "https://media.sume.com/artifacts/artf_demo/clean.mp4",
  "style": "black-outline",
  "design": {
    "colors": { "stroke": "#000000", "active": "#FFD400", "card": "rgba(0,0,0,0.55)" },
    "typography": { "stroke_width_px": 6 },
    "phrasing": { "max_words": 4 }
  }
}

How do you re-style without paying to transcribe again?

Pass source_caption_id instead of video_url and Sume reuses that caption's source video and its existing word timings, so no second speech-to-text runs. Billing is unchanged because a restyle is still a render ($0.20 per job for videos up to 60 seconds under the current fixed estimate, confirmed live in GET /v1/catalog).

To correct wording the way a phrase editor would, send words alongside source_caption_id. script_text, words, cues and segments are mutually exclusive, so choose one correction method per request. Our post on restyling without a second transcription walks through it.

What does Sume not do here?

Sume does not give you a visual editor, a template gallery that syncs across devices or live preview, and it does not export an SRT: standalone caption jobs reject SRT uploads and return a burned-in video. If you edit captions by hand in an editor and want a subtitle file out, an app like Captions fits better. If you need the same stroke and colours on 200 clips from a script, the design object is the more repeatable route.

Translated captions add one more reason to keep the look in a file. The same design can be reused across languages, but a longer translated line wraps earlier, so keep a per-language phrasing override (max_words or max_chars) next to the shared colours and let the validator reject anything out of range before it renders. A restyle of a clip you already captioned in Spanish does not touch the German job.

Test on a short clip first. Busy footage can swallow thin strokes, and a heavier stroke_width_px plus a card is the usual fix; the typography tokens post walks through each token.