Pick burn-in or SRT before export
Burn-in for the muted feed. Export SRT for the course player. Do not ship one object to both jobs.
Pick burn-in or SRT before you export. The feed wants pixels. The course wants a sidecar. They are not the same caption job, and exporting “captions” without naming the object is how a TikTok ships with a file nobody will play, or how an LMS gets a burned MP4 it cannot translate.
This is not a format explainer. Burned-in captions and SRT already have names. This week’s decision is which one you render for the file you are about to upload. Feed versus course. Ads versus the module. One radio, then export.
Closed captions are the sidecar idea: a track the player draws. Burn-in is the opposite: type composited into frames. Short-form almost never wants the sidecar. A muted autoplay feed will not enable a track the viewer never asked for. A course player, a long-form YouTube master, and a client who still restyles language do want the sidecar. Pick before the render, because burn-in is pixels. You do not “switch it off” later without a second job.
Feed is burn-in
Reels, TikTok, Shorts, paid social, and most UGC ads play muted until someone taps. The first sentence has to be in the picture. A .srt sitting next to the MP4 does nothing in that environment. The platform will encode the video. It will not look for your SubRip file. If you uploaded an SRT to YouTube Shorts or a Reel and assumed the words would appear, you captioned a player that is not there.
Burn-in is the feed object: transcription, timing, a look from caption styles, composited onto the frames. The viewer cannot hide it. The typeface is yours. Placement is yours, inside the safe area, not under the username stack. That is why the look is part of the edit rather than an accessibility toggle.
The cost of that object is a full caption render, 1× on the 21 basic presets or 2× on the 9 dynamic ones. You pay it because the feed will not show a sidecar. You do not pay it because burn-in is “better captions.” It is a different artefact.
If this week’s slate is eight vertical ads and two TikToks, the export radio is burn-in for all ten. Do not also produce SRTs “for later” unless someone asked. Unused sidecars are not a backup of burn-in. They are a transcript. Useful. Not the file the feed plays.
Watch the burn on a phone, muted, once, before you upload. If the first line sits under chrome, you did not finish. Move the preset position or recrop. Exporting a second time is cheaper than discovering the collision in ads manager.
Course is SRT
A module, a long-form YouTube video, a broadcast master, a client who will translate, a player with a CC button: those want a sidecar. SRT is the interchange: numbered cues, start and end times with a comma decimal, plain text. No styling. No position. That is the feature. The player draws the words. The viewer can switch them off. A second language is a second file, not a second encode of the picture.
Burning captions into a 40-minute course so that next quarter’s Spanish version requires a full re-composite is how teams learn this the expensive way. The picture should stay a picture. The words should stay text.
SDH — captions that also describe non-speech sound — usually lives in a sidecar too. Auto-transcribe will not invent [door slams]. If the brief is SDH, someone still authors those cues. The format is still SRT or WebVTT. The content of the cues is the extra work.
WebVTT can place a cue; SRT cannot. If the destination honours VTT positioning, author VTT. If the client asked for SRT, accept that placement falls back to the player default. Do not design a careful raise, convert to SRT, and expect the raise to survive. This week’s radio is still burn-in versus sidecar. The sidecar flavour is a second, smaller pick.
YouTube long-form is the mixed case. Closed captions help search and accessibility. Burn-in on a talking-head YouTube cut is optional and often noisy. If the same cut will also be clipped to Shorts, you have two exports: a master with a sidecar, and vertical clips with burn-in. That is two radios, two files. It is not one file asked to do both jobs.
Do not export one object for both
The failure mode is a single “captioned” MP4 sent to the social calendar and the LMS.
On the feed, a sidecar-only MP4 is an uncaptioned video. Mute viewers see a face. They do not see the line.
On the course, a burned MP4 is a picture with English (or whatever you burned) welded in. Translation means re-rendering. Accessibility settings cannot hide the type. A student who finds the burn distracting cannot turn it off. You chose pixels. Live with pixels, or do not burn.
If a client wants both — a social cut and a course master — render twice from the same transcript. Burn the social file. Export SRT (or VTT) beside the master. The transcript is the source of truth for words and times. The burn is a look. The sidecar is a track. Add captions to a video is the Versely path that transcribes and burns; sidecar conversion is a separate deliverable when the brief names a .srt.
Do not burn first “to be safe” and then hope you can recover a clean master. You can transcribe again. You cannot un-composite type without covering it or cutting around it. Picture lock, then pick the radio, then export.
Platform closed captions, when they exist, are not your SRT and not your burn. YouTube may auto-caption. TikTok may draw a default. Those are the platform’s type, the platform’s timing, the platform’s spelling of your product name. If the line matters, you either burn it or you upload a sidecar you actually checked. Hoping the platform will typeset the SKU is not a caption strategy.
The export radio for this week
Write the destination on the slate before anyone hits export:
| Destination this week | Object | Why |
|---|---|---|
| TikTok, Reels, Shorts, ads | Burn-in | Mute autoplay; no sidecar in the feed |
| Course / LMS / long-form master | SRT or VTT | Language swaps and CC without a re-encode |
| Same story in both places | Two files | One burn, one sidecar, one transcript |
If the week is only feed, you do not owe an SRT. If the week is only course, you do not owe a burn. If the week is both, you owe two exports, not a compromise file.
Safe area still applies to burn-in. It does not apply to SRT; the player places SRT. Do not spend an afternoon nudging SRT “position” in a format that cannot store position.
When the radio is burn-in, pick the preset before the full render — 1× basic or 2× dynamic is a different decision. When the radio is SRT, skip the look. Words and times. That is the whole sidecar.
If you need the studio to transcribe and burn the feed cut, sign up and export the object the destination can actually play. A caption the player never draws is not a caption.
FAQ
Can I upload an SRT to TikTok instead of burning?
Do not plan on it. The feed is a muted autoplay surface. Your SubRip file is not the soundtrack. Burn the line into the MP4 if the first sentence has to be readable with sound off.
We already burned the course. Can we still ship an SRT?
You can export a transcript as SRT for archives or translation, but the picture still has English welded in. For a clean multi-language master you need an unburned picture. Next module, pick sidecar first.
Is WebVTT the same radio as SRT?
Same radio: sidecar, not burn-in. VTT can carry placement and some styling; SRT cannot. If the player is a course site that honours VTT, prefer VTT. If the freelancer asked for SRT, send SRT and do not expect design.