Two features get described with the same word and behave nothing alike: exporting a subtitle track that exists, and producing one that does not.
Many public videos ship with a subtitle or caption track baked into the page. That track is a file — usually WebVTT or SRT — served alongside the media. Saving it is a copy operation: the text you get out is the text the publisher wrote, with their line breaks and their timing.
That is what saving available subtitles does. It is the useful case for study material, because a subtitle file can be searched in seconds, pasted into notes, compared against a human translation, or fed to a tool that builds a vocabulary list.
If a video has no track at all, no downloader will produce one. Turning speech into timed text requires a speech-recognition model, and that is a separate pipeline with its own accuracy profile, its own cost, and its own error modes — proper nouns and technical vocabulary being the usual casualties.
Worth knowing before you plan around it: a tool that quietly re-encodes or machine-transcribes is doing something other than what its label suggests. Keep the two verbs apart — save the track, transcribe the audio — and the rest of the workflow stays predictable.
Subtitle files reference the media by timestamps. If you later trim, join or re-encode the video, the track will drift out of sync unless the edit is applied to both. Save them as a pair, note the pairing in your index line, and keep the original file around even if you also keep a trimmed copy.
For lecture series and interview archives, the subtitle file is often the only part worth keeping long term. It is a few kilobytes, it survives format changes, and it is the thing you will actually search. Keep the video if the picture matters; keep the track either way.