Short answer: to download a video with subtitles from a website, you first identify how the captions are delivered. Three shapes cover almost everything: a separate text track (WebVTT or SRT) fetched by its own request, a subtitle track multiplexed inside the stream, or text rendered into the picture by the player itself. The first two survive a download; the third does not — and if it is burned in, nothing separates it afterwards. Multi-language audio is the same problem one layer down: it is usually several representations, and picking one is a choice you have to make explicitly.
Where captions live
| Delivery | How you can tell | What to do |
|---|---|---|
| WebVTT file per language | A .vtt request appears when the caption menu is opened; the element is a track element | Fetch that URL, keep it next to the video |
| SRT offered by the player | A “download captions” control, or an .srt request | Fetch and convert only if a device needs it |
| Subtitles inside the stream | ffprobe lists a subtitle stream in the same file | Keep the container that preserves it (MKV usually) |
| Closed captions as CEA-608/708 | A cc or cea track in the manifest | Extract to text with FFmpeg |
| Burned-in text | Captions appear even when the caption menu is off | Nothing to extract; they are pixels |
The WebVTT format itself is standardised as Web Video Text Tracks, and MDN documents the WebVTT API browsers use to consume it. The practical consequence: a caption track is a small text file with a URL, so capturing it is trivial once you see the request.
Downloading the caption file
Open the caption menu and let it fetch
Most players request subtitle files lazily, when a language is selected. With the Network panel filtered, click through the caption options: each language should produce a .vtt or .srt request. This is the same observation technique described in finding a video’s source URL.
Save each language with a matching name
Name them to the video: lecture-04.en.vtt, lecture-04.ar.vtt. A caption file whose name doesn’t match its video is a caption file you will not find later.
Test it against the video
Open the video in a player that accepts external subtitles and load the file. Check timing at the beginning and near the end — a mismatched frame rate or a truncated download shows up as drift.
Decide: sidecar or embedded
A sidecar keeps the video untouched and is the most portable for editing. Embedding is nicer for a single file you carry on a phone; FFmpeg can add a subtitle stream to MKV by copying both: ffmpeg -i video.mp4 -i subs.vtt -c copy -c:s srt out.mkv. See its formats documentation for which containers accept which subtitle codecs — MP4 is stricter than MKV here.
Download video and subtitles separately — when that’s the right answer
“Download video and subtitles separately” is usually the correct workflow, not a fallback:
- You keep the delivered video byte-for-byte — no re-encode, no generation loss (quality rules).
- Text tracks are editable: translate them, fix transcription errors, strip filler.
- They cost almost nothing to store, so keeping every language is practical.
- Burning captions into the picture is a re-encode of the whole video — the last resort, not a convenience.
And “save subtitles with downloaded online videos” is mostly a habit question: keep the caption file with the video, same folder, same stem, plus the source URL and retrieval date.
Audio tracks: extracting one, keeping several
Multi-language delivery is common on courses, broadcasts and dubbed content. Two distinct requests get confused constantly:
| You want | What it means | Method |
|---|---|---|
| One language only | Select a single audio representation and drop the rest | Choose it during capture, or map it when muxing |
| Audio only (a lecture as sound) | Extract the audio stream, discard video | ffmpeg -i in.mp4 -vn -c:a copy out.m4a |
| All languages preserved | Keep every audio track in one file | Remux into MKV, which tolerates multiple audio streams |
| Recompressed audio for size | Genuine transcode | Accept the loss, keep the master untouched |
Note the difference between “download audio track from a web video” (extract an existing stream, no encode, quality intact) and “convert to MP3” (transcode, loss). The first is a copy and should be your default; FFmpeg’s stream selection and container support are documented in the main manual and the extraction guide.
Separate audio is not the same problem
Adaptive delivery often ships audio as its own segment family, and fetching video alone produces a silent file. That is a missing-track case, not a caption case — handled in downloaded video has no audio, with the DASH equivalent in DASH manifests.
What to check in any tool before you trust it here
- Does it list subtitle/audio resources alongside video, or only “the video”?
- Does it default to copying streams rather than re-encoding? VidBrowser’s documented behaviour is FFmpeg-based detection with remux to MP4 or MKV — a copy operation, which is what you want for tracks and captions; anything that always converts should raise your suspicion.
- Does it keep the original text file if you don’t embed it? A tool that silently drops captions is worse than one you run twice.
- Licensing: caption files are often copyrighted text, sometimes human-translated. Keeping them for personal study is one thing; publishing them is another.
FAQ
Can I download a video with subtitles from a website?
When captions are delivered as a WebVTT or SRT request, or multiplexed in the stream, yes: capture that track and keep it with the file. Burned-in captions cannot be separated.
How do I download video and subtitles separately?
Open the caption menu so the player fetches each language, capture those URLs, and save them next to the video with matching names.
How do I keep subtitles after merging segments?
Either keep the caption file as a sidecar, or remux into MKV with the subtitle stream added by copy. MP4 accepts fewer subtitle formats than MKV.
Can I download just the audio from a web video?
Yes — extract the audio stream with copy rather than transcoding, which preserves quality and is much faster.
How do I keep multiple audio languages?
Fetch each representation you want and mux them into MKV, which carries several audio streams; then select the track at playback.
Are downloaded subtitle files editable?
Yes, they are plain text with timecodes. That is one reason sidecar files beat burning captions into the picture.