The quick answer

If your video needs to be understood with the sound off and on platforms that ignore sidecar files (Instagram, TikTok, most Shorts feeds), burn the captions in. If your video lives on YouTube or a course platform where viewers can toggle a CC button and you want indexable text plus translation options, upload a separate caption or subtitle file instead. Most working creators end up doing both: an SRT or WebVTT file for the long-form master, and a burned-in version for every short-form clip.

That two-format habit is the heart of a modern caption workflow. The rest of this guide explains why, and how to choose for each piece of content you ship.

What “burned-in” actually means

Burned-in captions (also called open captions) are text that has been rendered into the video frames themselves. They are part of the picture. A viewer cannot turn them off, the platform cannot strip them out, and there is no separate file travelling alongside the video.

A sidecar caption file — usually .srt, .vtt, .scc, .ttml or .imsc — is a small text document that sits next to the video. The player reads the file, the viewer presses a CC button, and the text appears. YouTube, Vimeo, LinkedIn, X and Facebook all support sidecar caption files in some form. Instagram, TikTok and most short-form feeds do not. Once you know that split, most of the caption decisions get easier.

The distinction matters for two reasons that are easy to miss:

  • What the text contains. Captions are written for someone who cannot hear the audio, so they include dialogue, speaker labels and non-speech cues like [door slams] or [upbeat music]. Subtitles are written for someone who can hear the audio but needs the words in another language.
  • Whether the viewer controls it. Closed or sidecar captions are toggleable. Open or burned-in captions are not.

Keep those two axes apart and the labels stop being confusing.

Why the choice matters for a creator workflow

Captions are not a finishing touch. They change four things about how your video performs:

  1. Accessibility for deaf and hard-of-hearing viewers. A same-language text track that includes non-speech audio is what an accessibility checklist actually requires. Plain subtitles in a foreign language are not a substitute.
  2. Silent autoplay behaviour. Most short-form feeds play without sound. If the text is not on the picture, viewers scroll past.
  3. Search indexing. YouTube and Vimeo use uploaded caption files as a primary text source for video search. Auto-generated captions are a fallback that platforms do not trust as much.
  4. Repurposing. Every long-form video tends to become a handful of vertical clips. The question “do I burn captions in or attach a file?” gets asked once per clip, not once per channel.

A simple rule of thumb: closed captions protect accessibility and SEO, open captions protect silent viewing and reach. Pick the format that protects the outcome you actually need for that specific export.

When burned-in captions are the right call

Burn them in when any of these are true:

  • The video is going to Instagram, TikTok, YouTube Shorts or LinkedIn feed video — places where sidecar caption files are not reliably honoured.
  • The video will be screened in a public space (events, trade shows, lobby loops) where there is no CC button.
  • The video will be downloaded and shared as a single file (email, WhatsApp, Discord, course download bundles).
  • You are clipping a longer video and the clip is shorter than about 60 seconds. The cost of burned-in text on a short clip is small and the silent-watch benefit is large.
  • Your audience is known to watch muted first and unmute only if the text pulls them in.

Trade-offs to accept when you burn in:

  • You cannot edit the wording without re-rendering the video. A typo means a re-export.
  • Viewers cannot translate your text into their own language using the platform’s built-in tools, because there is no sidecar file to swap.
  • You lose styling control on the platform side — whatever font, size and background you bake in is what everyone sees.
  • File size grows slightly because the rendered text takes up encoded pixels. For a 60-second vertical clip this is usually negligible; for a two-hour podcast master it can add up.

When a separate caption or subtitle file is the right call

Upload a sidecar file when any of these are true:

  • The video lives on YouTube, Vimeo or a learning platform that supports a CC toggle and language tracks.
  • You want the platform’s own translation features to work on your text. YouTube, for example, can auto-translate an uploaded caption track into other languages, which it cannot do from burned-in pixels.
  • You plan to keep editing the transcript after publishing — fixing names, adding speaker labels, correcting jargon. A sidecar file can be re-uploaded and replaced without re-rendering the video.
  • You want one source of truth. The same SRT can be the master for YouTube, the master for Vimeo, and the basis for translated subtitle tracks, with no extra rendering step.
  • The video is long-form (interviews, podcasts, lectures) where fixing a typo in burned-in text is genuinely expensive.

Trade-offs to accept with sidecar files:

  • The text only appears if the viewer turns it on, and many viewers never do.
  • On feeds that auto-play muted, sidecar captions often do not show at all.
  • You have to manage the file: version it, rename it, upload it again when you re-upload the video, and remember it exists.

The format question: SRT, VTT, SCC, TTML

Most creators only need to know two formats.

  • SRT (SubRip) is the workhorse. Plain text, timecodes in hours:minutes:seconds,milliseconds, plain text lines underneath. YouTube, Facebook, LinkedIn and Vimeo accept it. A “Google/YouTube variant” of SRT exists with slightly stricter formatting, so if you are exporting from a tool, check that the target platform is selected.
  • WebVTT is the web-native successor to SRT. It adds styling hooks (positioning, colour, background) that SRT does not support. HTML5 video players, Vimeo and some course platforms prefer VTT when you want any styling control beyond font and size.

SCC and TTML are broadcast and streaming-platform formats. You only need them if you are delivering to a broadcaster, a TV app or an enterprise LMS with strict specs. For YouTube, Vimeo, podcast clips, courses and social, SRT is enough.

A practical note: if your editor exports an MP4 with captions embedded in the video track itself, that is a third option — sometimes called “embedded” or “muxed” captions. It looks identical to burned-in to a viewer, but the text is technically a track inside the container. Some editors prefer this for masters because the text and picture stay locked together. The trade-offs are the same as burned-in.

Platform-by-platform in plain English

  • YouTube: Accepts uploaded SRT and WebVTT files for any video. Auto-captions exist and can be edited, but an uploaded file gives you cleaner text, better indexing and translation options. Burned-in captions are rarely needed because the CC button works well.
  • Vimeo: Accepts uploaded caption files, supports multiple language tracks, and shows a clear CC toggle. Same logic as YouTube — upload a file, do not burn in.
  • Facebook: Accepts SRT files for longer videos. Reels do not reliably honour sidecar files, so burn captions in for Reels.
  • LinkedIn: Accepts SRT for native uploads, but many viewers will not see the CC toggle. Consider burning in for feed video.
  • X (Twitter): Accepts SRT. On iOS and Android, captions appear automatically when the sound is off; on the web, viewers toggle CC. Behaviour varies by client, so a burned-in fallback is reasonable for important posts.
  • Instagram (feed, Reels, Stories): Does not accept sidecar caption files in a reliable way. Burn captions in, or use the platform’s own caption sticker / auto-caption tool.
  • TikTok: No reliable sidecar upload. Burn captions in, or use TikTok’s auto-captions and edit them.
  • Shorts: Behaves more like TikTok than YouTube. Burn captions in for the clip itself; keep the SRT on the long-form master.

The pattern is clear. Long-form, owned platforms: sidecar files. Short-form feeds and social: burned-in.

Accessibility: which format actually satisfies a checklist?

A real caption track — same language, dialogue plus non-speech cues, accurate to roughly 99 percent, timed within half a second of the audio, displayed at a readable speed — satisfies an accessibility checklist whether it is delivered as a sidecar file or burned in.

The difference is what happens next. With a sidecar file, the viewer chooses. With burned-in text, the captions are always there. From a pure accessibility standpoint, burned-in captions are arguably safer because they cannot be switched off by accident. But you give up translation and you make editing expensive.

For most independent creators, the practical answer is: produce one clean, accurate caption track. Deliver it as a sidecar file for long-form and any place a CC toggle exists, and burn a copy of that same text into every short-form clip.

Auto-generated captions from any platform are a starting point, not a deliverable. They typically need a manual pass for names, jargon, speaker labels and non-speech cues before they meet any reasonable accuracy standard.

SEO and indexing: what a sidecar file really does for you

When you upload an SRT to YouTube, the platform ingests the text and uses it for video search the same way it uses your title, description and tags. Auto-captions are used as a weaker fallback, and burned-in text is not indexed at all because the platform can only see pixels.

This is the single biggest reason to keep a sidecar file for your long-form master even if you also burn captions into short-form clips. The SRT is doing SEO work for you while the burned-in version is doing silent-watch work for you. They are not substitutes.

Repurposing workflow that does not double your work

The most common caption mistake is rebuilding text three times for one video. Here is a workflow that avoids that.

  1. Edit the long-form master first. Decide on the final transcript. Fix names, add speaker labels, note sound effects.
  2. Export one clean SRT. This is your source of truth. Version it with the project name and date.
  3. Upload the SRT to YouTube or Vimeo alongside the long-form video. Do not rely on auto-captions.
  4. Cut your short-form clips from the master. Use the same SRT as the basis for burned-in captions. Most video editors (Premiere Pro, DaVinci Resolve, Final Cut, CapCut, Descript) can import an SRT and either export it as sidecar text or render it into the picture.
  5. Render two versions of each short clip: one with burned-in captions for TikTok, Reels and Shorts feeds, and one clean version without burned-in text for cases where the platform’s own caption tool will be used instead.
  6. Translate once, deliver many times. If you commission subtitles in Spanish, Japanese, etc., export them as additional SRT or VTT files for YouTube’s multi-language track feature. Do not bake translations into the video.

This is also where a tool like Descript, Simon Says, Whisper-based workflows or a human captioner earns its keep: you make the text right once and push it everywhere.

Styling control: how much can you actually change?

With a sidecar file, your styling options depend on the player. YouTube lets you pick font family, size, colour and background colour from a fixed menu. Vimeo offers similar controls. WebVTT gives you more, including positioning, but only on players that honour the spec.

With burned-in captions, you control everything because you are rendering the text yourself. Font, weight, drop shadow, background box, animation, word-by-word highlighting — all of it. This is why creators who care about brand style on social almost always burn captions in for short-form. The price is that you have to design it once per template and re-render when you change the wording.

A practical compromise: keep a small set of caption presets in your editor — one clean lower-third look, one bold centred look, one karaoke-style word highlight — and apply them per platform.

File size and export trade-offs

Burned-in captions add pixels to every frame. For a 30 to 90 second vertical clip, the file size increase is small and worth it. For a 10 to 60 minute interview, it can add real weight to the export, and you lose the ability to swap captions without re-encoding the video.

Sidecar files are tiny — usually under 50 KB even for a two-hour podcast — and they add essentially nothing to the upload size.

If you are exporting an MP4 master for archival, keep the SRT as a separate file in the same folder. If you are exporting a clip for social, bake the text in.

A short checklist before you export

  • Have you exported a clean SRT from the final edit, with names and non-speech cues corrected?
  • Does the long-form destination support sidecar uploads? If yes, upload the SRT. Do not rely on auto-captions.
  • Is the clip going somewhere that ignores sidecar files (TikTok, Reels, Shorts, Stories)? Burn captions in.
  • Have you burned in a version that you can swap if you need to fix a typo? Keep the editable project; do not throw it away after export.
  • If you publish in multiple languages, are the translated tracks delivered as sidecar files, not baked into the video?

FAQ

Do burned-in captions count as accessible? They can. The text is always visible, which removes the “viewer forgot to turn captions on” failure mode. The trade-off is no toggle, no easy translation and expensive edits. For most workflows, an accurate SRT uploaded as a sidecar file plus a burned-in version for short-form clips is a stronger answer.

Should I trust auto-generated captions? Treat them as a first draft. Expect to do a manual pass for names, jargon, speaker labels and non-speech cues before you publish, especially for any content that needs to meet a basic accuracy bar.

Can I upload one SRT and have it work everywhere? Usually yes. YouTube, Vimeo, Facebook and LinkedIn all accept SRT. The main exception is Instagram-style feeds and TikTok, which is exactly why you also burn captions into those clips.

What if I want animated, word-by-word captions? Burn them in. Sidecar formats either do not support that level of styling or only support it on players that fully honour WebVTT, which is a minority. Animated captions are a short-form, burned-in job.

Do burned-in captions hurt SEO? They do not help, because the platform only sees pixels. This is the main reason to keep an SRT for your long-form master even when you burn captions into every clip.

Is an MP4 with embedded captions the same as burned-in? For viewers, yes — the text is part of what they watch. For delivery, no — embedded captions live as a track inside the file and can sometimes be toggled by a player that understands them. If you are not sure whether your audience’s player will honour that toggle, treat it as burned-in and assume it cannot be switched off.

Sources