Keeping video, audio and subtitles — a field guide

Most download workflows fail at the first step, not the last: people decide which tool to use before deciding what they actually need to keep.

A link to a public video is not a single object. It is a page that offers a picture track, usually an audio track, sometimes a subtitle track, and always a set of permissions. Four decisions follow from that, and each has a different right answer.

1. Pick the artefact first

• A talk you will listen to on a commute. Keep the audio. An hour of speech is roughly 50–100 MB; the same hour as video is several times larger and you were never going to look at the slides.
• Language material or a lecture you need to quote. Keep the subtitle file. It is searchable, comparable against your own notes, and small enough to read.
• Reference footage where the picture carries the information. This is the only case that needs the full video.

Choosing badly is not free. It costs you a large file you never open, or the discovery — six months later — that the part you needed was the transcript.

2. The browser route covers the ordinary case

For a public page or a direct media URL, a browser tool with no installation is enough: paste the address, wait for the available formats to appear, choose one, save the finished file. Keep Video's online downloader works this way, with no account required.

One honest limit: available formats follow the source. A 480p stream cannot be exported as 1080p, and any tool that claims otherwise is re-encoding the picture and making it worse in the process.

3. The subtitle track is a separate asset

When a video already ships subtitles, that track can be saved on its own and reused: searched, translated, diffed against a transcript, or attached to a set of notes. What a downloader cannot do is invent subtitles for a video that has none — that is a transcription job, not a download. The distinction matters before you build a workflow around it; there is a short note on saving available subtitles that says the same thing.

4. Some material only plays inside your own session

A server-side tool sees what an anonymous visitor sees. If the material is only visible to a signed-in account, the download has to happen on a machine that holds that session. That is the case the desktop app exists for: it runs locally on macOS, Windows and Linux, adds download queues and local file management, and keeps control of runtime dependencies on your side of the connection.

5. Name files for the person you will be in six months

A folder full of video_final_2.mp4 is not an archive. One line per item, in a plain text file next to the media, is enough:

2026-03-14 | series-02-ep05 | source URL | audio + subtitles | cited in chapter 4

Date, filename, source, what you kept, why. Ten seconds per item, and the difference between a reference library and a pile of anonymous files.

6. Permission is part of the workflow

Saving your own uploads, material you are licensed to keep, and content prepared for permitted offline study are the intended uses. A downloader does not change the copyright status of what you point it at — that part is on the person holding the mouse.