siphon checking server…

Download video and audio

Link to download

Quality

Subtitles

None In the video Separate files

Subtitle languages

Comma-separated. Auto-generated captions are included, since most videos have nothing else.

On this device subtitles are embedded in the video, when the link offers any — a plain file or a scraped page has none to offer. Separate .srt files need your own server.

Downloads

How this works

Paste a link. A video page, a direct file, a stream. Tap Paste, or share a link into this app.
Pick a quality, tap Download. Fetching, merging and converting happen on this device. Nothing is uploaded.
Save it. The file lands in your downloads, and the row stays here until you clear it.

On a computer: drag Send to siphon to the bookmarks bar, and on any video page one click opens this app with that page's link. A link dropped anywhere on this page, or a pasted list of them, is taken too — several become one row each.

YouTube

A free relay — one click on Cloudflare, then paste the address it gives you into settings. Deploy the relay

The bridge — a userscript; nothing leaves this device. Install the bridge

Your own server — the most capable: playlists, subtitles, your cookies. docker run -d -p 8000:8000 -v siphon:/tmp/siphon ghcr.io/maxgfr/siphon

Running this site? Set the repository variable SIPHON_RELAY_URL to your relay once, and every visitor gets it by default.

Download only what you have the right to download, and respect each site's terms.

Where downloads come from

Paste a link and it downloads on this device. Some sites — YouTube above all — refuse a web page, and for those this app needs a helper. Give it one address; it works out what that address is.

Helper — optional

Any of these, and it is recognised on its own: your own siphon server — the server/ folder, one docker run. With ffmpeg it does everything: every site yt-dlp knows, playlists, subtitles, your YouTube cookies. Without ffmpeg it only resolves links and this device downloads and converts. A cobalt instance, for what this device cannot read. An Invidious or Piped instance, for YouTube. A relay — relay/, a free Cloudflare Worker — which adds the missing header and nothing else. Leave it empty and this device still handles direct files, HLS and pages that allow it.

Public instances are someone else's server: whichever one is set is named here and on the main screen, it sees every link it is asked about, and clearing the field undoes it. The list beside this app is measured daily, from a machine with real network, by asking each instance exactly what this page asks; only the ones that answered are offered.

Access key — if the helper asks for one

On your phone

Advanced — yt-dlp options

Remove sponsor segments

Sponsors, self-promotion and "like and subscribe" are cut out of YouTube videos, from the community's SponsorBlock data.

Clip — only this part

As 1:23 or 01:02:03. Leave both empty for the whole video.

Speed limit — optional

Bytes per second, as 500K or 2M, so a download does not eat the connection.

YouTube client

Which of YouTube's clients to try first. The others still follow when it is turned away.

YouTube sign-in

YouTube turns anonymous downloads away with "Sign in to confirm you're not a bot". Your server already retries with several of YouTube's own clients, and that clears most of it. When it does not, your own session does — export cookies.txt with a Netscape-format cookie extension and upload it here.

Upload cookies.txt

This is a logged-in session. Only upload it to a server you control, and use a throwaway account if you would rather not hand over your main one.

The bridge. A userscript that lets this page fetch any public address, with no server anywhere — see bridge/ in the project. Once installed it is used automatically, ahead of any helper.

Not checked yet