There are two ways to ship captions, and they are more different than they look. Burned-in captions are pixels: the words are drawn into the picture at export, so they appear on every player, in every embed, at exactly the size and position you chose — and they can never be changed, hidden or translated again. An .srt file is the opposite: plain text with timestamps, carried beside the video, rendered by whatever plays it. The player decides the font; the viewer decides whether the captions are on at all. Neither is simply better. The question is who you want in control after publishing: you, permanently, in exchange for flexibility — or the viewer, in exchange for trusting the platform to display the track at all.
What each platform does with them settles most cases. YouTube treats an uploaded subtitle file as first-class: viewers toggle it, resize it, machine-translate it, and the text is searchable — which a burned caption never is, because to a search engine pixels are not words. LinkedIn also accepts a subtitle file on uploaded video. The vertical feeds are the other world: TikTok and Instagram autoplay muted, their own caption tools vary, and a separate track is not guaranteed to be shown the way you intended — which is why burned-in captions dominate there, and why they suit those platforms. So the practical answer is usually both: burn the captions for the vertical feeds, and give YouTube a clean video plus an .srt so the viewer keeps their controls.
Accessibility argues for the file. A deaf viewer with their own caption preferences — larger text, higher contrast — can apply them to a track and not to pixels; a blind viewer’s tooling can read a track aloud; anyone can machine-translate one. Burned-in text is locked at whatever size looked right to you on the day, which on a small screen may be nobody’s idea of readable. Editing later argues the same way. A typo in an .srt is a text edit; a typo burned into the picture means re-exporting and re-uploading the whole video, and on most platforms that means losing its comments and its view count. If you burn, proofread like it is print — because it is.
Whichever form you choose, captions are only as good as their words and their clocks. Talk2Camera takes the words from the script you actually read, so names, jargon and anything a speech recogniser has never heard come out spelled the way you wrote them, not the way a transcription guessed. The script is saved beside the recording, which means this still works when you open a take next week, not only in the minutes after shooting it. Timing survives editing too: cut a stretch of the take and the timeline closes up and the captions retime to match, so the .srt you export describes the video you actually published rather than the one you recorded. Both exports come from the same captions — burned into the picture for the vertical feeds, or a separate .srt for the platforms that deserve one.