Twitch Copyright & Privacy: What's Legal When You Download (UrhG & GDPR)

Updated: 2026-07-06 · Written by the vodfetch founder

Who owns a Twitch stream, and what are you actually allowed to download? A plain-English overview of the copyright (UrhG) and data-protection (GDPR/DSGVO) questions around saving Twitch videos — the private-copy concept, personal data in streams and chat, and where redistribution crosses a line. This is general information, not legal advice.

First: this is general information, not legal advice

Copyright and data-protection law is genuinely complicated, varies by country, and turns on the specifics of your situation. Nothing here is legal advice — it's a plain-English overview to help you understand the questions involved. For anything that actually matters, talk to a qualified lawyer (in Germany, a Rechtsanwalt).

With that said, most of the everyday confusion comes down to a few clear ideas, and understanding them helps you stay on the responsible side of the line.

Who owns the content in a Twitch stream?

A stream is a stack of different rights. The streamer owns their own footage, commentary and camera. The game publisher owns the game's visuals and audio. Any background music belongs to its rights holders. And Twitch has its own Terms of Service governing what you can do on the platform. Downloading a VOD doesn't transfer any of those rights to you — you're making a copy of someone else's protected work.

That's the core principle in most jurisdictions, including under German copyright law (Urheberrecht, UrhG): creative content is protected by default, and the copyright holder decides how it may be used and shared.

The private-copy idea (Privatkopie) under German law

German copyright law recognises a limited private-copy exception (§53 UrhG): individuals may generally make copies of a work for their own private, non-commercial use, within limits. This is the concept people usually mean when they say downloading "for personal use" is treated differently from redistributing.

But it's narrow and conditional — it does not permit copying from an obviously unlawful source, it doesn't allow circumventing effective technical protection measures, and it never extends to publishing, re-uploading or monetising the copy. Whether any specific download qualifies is exactly the kind of detail a lawyer, not a blog post, should judge.

GDPR / DSGVO: streams and chat contain personal data

There's a second layer people forget: a Twitch VOD and its chat can contain personal data — faces, voices, usernames, messages. Under the EU's GDPR (in Germany, the DSGVO), processing other people's personal data has rules, especially if you go beyond purely private use and start publishing or sharing it.

For keeping a private copy for yourself, everyday private-use exemptions usually apply. The moment you republish a VOD or chat log containing identifiable people, you're processing their data in a way that can trigger obligations — another reason redistribution is where the real risk lives.

Twitch's Terms of Service are separate from the law

It's worth separating two things: what the law allows, and what Twitch's Terms of Service allow. They're not the same. Twitch's ToS is a contract between you and Twitch; something can be permitted by law but still against the ToS, or vice versa. Downloading tools generally operate in the space of personal use, but you should read and respect Twitch's current Terms.

The honest summary: your own content and content you have permission to save are the safe cases; other people's content for private, personal use sits in the nuanced private-copy area; and redistributing or monetising anything you don't own is where problems clearly begin.

Practical, responsible guidance

In plain terms: downloading your own VODs to archive them is the clearest case. Saving a public clip or VOD for genuinely private, personal use is the grey area the private-copy concept is about. Re-uploading, streaming publicly, or monetising someone else's content without permission is where you're most likely to be in the wrong — both under copyright and, if it contains personal data, under the DSGVO.

A tool being privacy-respecting doesn't change what you're allowed to download — that's on the user. vodfetch keeps no account and stores no downloads, and we're transparent about exactly how it works, but the responsibility to only save content you own or are permitted to save stays with you. When in doubt, ask a lawyer.

Access controls the tool does not bypass

There is no login step in vodfetch and no user token. Its GraphQL requests go straight to gql.twitch.tv with Twitch's public web Client-ID — the same anonymous identity a logged-out visitor has. Ask it for a subscriber-only or private past broadcast and Twitch returns no playback access token at all, so the run stops with "VOD unavailable (deleted, private or sub-only)". If a request gets past that stage and the playlist server refuses instead, the message is "Usher 403 (sub-only/expired)".

That wall is Twitch's own access control working, not a bug to route around. A tool holding no subscriber credential cannot unlock paywalled content; it fails. Note that the private-copy exception in §53 UrhG explicitly does not cover circumventing effective technical protection measures — and note equally that failing to bypass a paywall says nothing about whether everything that does download is lawful.

Deletion is the other wall. Past broadcasts disappear after 7 days, 14 for Affiliates, and 60 for Partners (and for channels with Prime or Turbo). Once Twitch stops serving the video there is no token to issue and you get the same "VOD unavailable" message, because there is nothing left at the source to fetch. Clips and highlights are off that clock, though highlights and uploads have shared a 100-hour storage cap per channel since 19 April 2025, announced in February 2025 — which leaves clips as the durable form. No downloader can retrieve a deleted VOD, whatever a competing tool's marketing implies.

Where your own data goes while you download

The sections above deal with other people's data; here is your side of it. Metadata queries leave your browser for gql.twitch.tv directly, so Twitch sees a request from your IP much as it would from any logged-out viewer. The media itself — playlists, HLS segments, clip MP4s — is relayed through a stateless Cloudflare Worker, purely because those hosts send no CORS headers and the browser would otherwise refuse to read them. Those relayed requests carry a short-lived signed ticket the page fetches from vodfetch first, which exists so the relay cannot be used as a free download backend by other sites. The Worker keeps nothing.

The finished file never exists on a server; segments are assembled inside the page. In browsers with the File System Access API the tool opens a save dialog before it starts and streams bytes straight into the file you picked. If that handle fails it falls back to memory and says so: "Could not stream to disk — buffering in memory". Browsers without the API buffer the whole file in RAM with no message at all, which is why on those browsers a download estimated above roughly 700 MB asks you to confirm first. The only local trace the site keeps is a twdl_recent entry in localStorage holding the last eight filenames and dates, which clearing site data removes.

The honest caveat: the site is funded by Google AdSense, so Google's ad and consent tooling runs on the pages as it does on most ad-supported sites. In the EEA and UK a certified consent message handles that, and the footer link labelled Cookie settings reopens it so you can change or withdraw what you agreed to — the link stays hidden where no consent message runs, because there is nothing to withdraw. The ad units are standard and each is labelled as advertising; there are no pop-ups and no fake download buttons. Downloads behave identically whichever way you answer.

The chat export is the file that actually contains personal data

Of everything the tool produces, the chat log is the one output that is mostly other people's data. It saves as "<vodid>_chat.txt", plain text, one line per message: a timestamp in brackets, the display name, a colon, the message. The timestamp is the offset into the broadcast and drops the hour until the VOD passes one, so lines read "[07:41] Name: text" early on and "[2:07:41] Name: text" later. Messages come from Twitch's public GraphQL comments field in pages of 100, and the paging loop stops after 3,000 pages — an upper bound of 300,000 lines, a searchable record of who said what.

That changes how you should handle it. Keeping a copy to find a moment again is ordinary private use; posting the raw file republishes identifiable people's messages in bulk, which is exactly the processing the DSGVO cares about. If you need to show chat publicly, quote the handful of lines you actually need, or replace the names. The chapters export, "<vodid>_chapters.txt", is far safer to share: each line is a timestamp and a chapter description from the broadcast itself, with no usernames in it.

Two practical notes. Both buttons appear only for past broadcasts — there is no chat or chapter export for clips or live streams — and the chapter button stays hidden on top of that when the VOD carries no markers. A replay whose comments are gone or unreachable ends with "No chat available for this VOD". And because the result is plain text on your own disk, deleting it when you are finished is the whole of data minimisation here.

Muted audio, music rights, and what a silent section tells you

If part of your download is silent, that is not a fault in the transfer. Twitch mutes VOD sections its system flags for copyrighted music, and the muting is baked into the audio Twitch serves — every downloader, including this one, receives the muted version. No quality setting and no format restores it, because the unmuted audio was never sent to you. The video track is untouched; only the flagged span loses its sound.

Read the mute as information. It means an automated rights system has already objected to the music in that section. Splicing the original track back in for a public upload steps directly over that objection, and platforms run their own matching regardless of the legal position — YouTube's Content ID commonly acts on a re-upload whether or not anyone files a complaint. Exporting audio only changes nothing here: it saves as .m4a, or .aac when the MP4 converter is not in play, never MP3, from the same stream.

One more thing to check before publishing anything cut from a VOD. Trimming keeps whole segments: the selection expands outward to segment boundaries at both ends, so the saved file starts slightly before and ends slightly after the times you typed. If you trimmed specifically to remove something — a name, a face, a song — play the first and last seconds of the result before sharing it, because a few extra seconds may have travelled with it.

Provenance: your filename is already a source record

Leave the filename box empty and the tool builds a name from metadata: broadcaster display name, publish date as YYYY-MM-DD, the video title, then the quality label with the bitrate part after the middle dot dropped, joined by underscores. Characters Windows rejects — backslash, slash, colon, asterisk, question mark, double quote, angle brackets, pipe — become underscores, and the whole string is capped at 120 characters. Rename that to "vod1.mp4" and you have discarded, for no gain, the only record of whose work the file is.

Keep the generated name, or store the source URL beside the file. Provenance is what lets you answer "whose is this?" a year later, and every use that leans on attribution needs it: German law treats quotation (§51 UrhG) separately from private copying, it is purpose-bound, and it expects the source to be named. Whether your particular commentary qualifies is a question for a lawyer; having the source recorded is a precondition under any answer you get.

For clips the stable citation is the original twitch.tv URL and its slug. Note what the metadata does and does not tell you: the tool reads the broadcaster's display name, because that is the only name the clip query asks Twitch for — not the viewer who created the clip, and on Twitch those are frequently different people. Clips always save as MP4, so the format switch cannot change what the file is, and the filename is the only origin marker it carries.

How to download a Twitch video

1. 1

   Prefer your own content

   Downloading and archiving your own past broadcasts is the clearest, safest case — you own the rights to your footage.

2. 2

   Keep other people's content private

   If you save someone else's public VOD or clip, keep it for genuinely private, personal use. Don't treat a private copy as a licence to publish.

3. 3

   Never re-upload or monetise what you don't own

   Publishing, re-streaming or making money from another creator's content without permission is where copyright — and, for personal data, the DSGVO — problems clearly start.

4. 4

   When it matters, ask a lawyer

   For any real decision, get advice from a qualified lawyer (Rechtsanwalt). This guide is general information, not legal advice.

Frequently asked questions

Ready to download? Use the free Twitch Video Downloader.

← Back to the Twitch Downloader