Platform captions vs burn-in on mute

TikTok, Instagram, and YouTube auto-captions are a player overlay. The viewer can ignore them. The platform skins them. The algorithm does not treat them as your art. Burned-in captions are in the file. Mute autoplay only shows pixels. Do not ship a silent cut and trust the app to typeset the hook.

A toggle is not a soundtrack

Closed captions, native auto-captions, “CC,” the in-app transcript bubble — all of these are drawn at playback. Someone has to enable them, or an account default has to enable them, and either can change. The typeface is the platform’s. The placement is the platform’s. The timing is the platform’s best guess, which will not match your cut if you trimmed after upload.

Mute autoplay does not wait. The impression is a silent frame. If the claim is only in a track the viewer never asked for, the claim is not in the impression. That is the whole mute-autoplay problem, and it is why short-form moved the words into the picture.

Platform captions also stack. A viewer with auto-captions on, watching your burned file, sees two transcripts. Yours, designed. Theirs, default, often on top of yours, often late. You cannot turn theirs off for them. You can keep your block compact and out of the densest chrome so a double-caption session is ugly instead of unreadable.

Search and accessibility teams like sidecar text because machines can read it. Feeds do not play that text. YouTube long-form can ingest an SRT and index it. Reels and TikTok will not make your .srt the For You soundtrack. Closed captions remain the right object for the room with a player. They are the wrong object as the only caption on a muted 9:16.

What each app is actually drawing

TikTok. Auto-captions are an in-app overlay. They are not your font, not your karaoke preset, not your safe-area layout. They can be useful as a fallback for viewers who enable them on other people’s silent videos. On your video, if the hook matters, burn it. TikTok will still draw username, caption field, and the right rail on top of whatever you burned. Plan for that. Do not also plan for TikTok to “use the SRT.”

Instagram Reels. Native captions and stickers are not a timed transcript you control in review. Paid Reels add a CTA. The mute default remains. Burn speech. Author the legal line as overlay. Treat Instagram’s own caption UI as chrome, not as delivery.

YouTube Shorts. Same mute-first vertical feed. YouTube may offer auto-captions as a player feature; it will also draw its own labels. As of 27 May 2026 the YouTube Blog places photoreal / meaningfully AI-altered labels as an overlay on the Short (Improving AI labels for viewers and creators). That overlay is not a caption, but it occupies caption space. Labels alone do not affect recommendations or monetization eligibility, per that post. They do cover type. Burn-in still has to clear the chrome and that badge.

YouTube long-form. Here the platform caption is a real destination. Upload an SRT or use Studio. Viewers expect a CC button. You can burn-in as well if the look is the show. You should not rely on burn-in alone if you need a second language or a patch without a re-render.

None of these overlays are word-highlight captions as you designed them. Karaoke is a burn look (Snap, Reveal; VEED DYNAMIC). The app’s highlighter, if it has one, is a different clock on a different transcript.

Burn-in is the file

Speech is transcribed or taken from an SRT you edited. Style is applied. Glyphs are composited. After that, re-encode and re-upload still show the words. That is the property you are buying.

It costs you localisation (a new language is a new render) and it costs you a recut (pixels do not ripple). Those costs are why you lock the picture first, fix names and numbers second, and only then burn. They are not a reason to skip the burn on a Reel.

BASIC (21 at 1×) is the quiet file: ads, courses, anything a reviewer will stare at. DYNAMIC (9 at 2×) is the hook file: organic spoken phrase, mute-pass readable. Neither tier is “platform captions.” Both are pixels.

Placement is part of the file. Safe area is a snapshot of chrome. Recheck. Bottom ~200px on a 1080×1920 export is a working keep-out, not a contract. Shop tabs and AI overlays grow the keep-out.

If you need both objects — and many teams do — render the burned short and keep the sidecar for the archive. Convert with the VTT to SRT converter. Do not rename. Do not attach the sidecar to the TikTok and skip the burn.

Mute is the judge

Play the export in the app’s draft preview, sound off, phone size.

• If the hook is readable with their auto-captions disabled, you have a file.
• If the hook only appears after you enable CC, you have a sidecar habit on a feed.
• If both your burn and their overlay are on, and the frame is soup, your block is too large or too low. Shrink the job (two lines, one graphic), not the font into a paragraph.

The AI caption generator and add captions to a video are the burn path: 165 language codes, VEED BASIC and DYNAMIC. They will not enable TikTok’s overlay for you, and they will not disable it. They put words in the MP4.

Platform captions are a player feature. Burn-in is the cut. Mute autoplay only sees the cut.

Burn the mute version in Versely.

FAQ

If my account has auto-captions on, do I still need burn-in?

Yes, if you care how the hook looks and whether it appears for viewers who do not have that default. Their setting is not your art. Mute autoplay still starts silent.

Do platform captions help SEO on TikTok the way an SRT helps YouTube?

Do not treat TikTok’s overlay as a substitute for words in the picture. YouTube long-form can index a sidecar. Short-form still needs pixels for the mute impression. Keep an SRT for the destinations that ingest one.

Can I use platform karaoke instead of Snap or Glass?

Not as a controlled look. Word-highlight is a burn preset on your transcript. The app’s highlighter is a different transcript, a different font, and a toggle. If karaoke is the creative, burn it.