Twitch Chat DownloaderExport a Twitch VOD's or clip's chat replay as text, JSON or a subtitle track.

Every message becomes one line in the form [12:04] ViewerName: message text. The number in brackets is the message's offset into the VOD rather than a wall clock, so it lines up with the position shown on the Twitch player's scrubber. Below one hour the prefix is minutes and seconds only, as in [07:03]; past the hour mark an hour figure is added, as in [2:17:30]. Lines run in order from offset zero, joined by plain newlines, with no header row and no summary at the end. Any trim range you set for the video is ignored here: the chat file always covers the whole broadcast.

Exactly three things are read from Twitch for each message: the offset in seconds, the commenter's display name, and the text of every fragment of the message. Emotes therefore arrive as the words that were actually typed, such as Kappa or PogChamp, and the file carries no images, no badges, no name colours and no user or message IDs. Where a message has no commenter attached, the name position shows a single question mark. The text is written as UTF-8 with no byte-order mark, so non-Latin names and messages survive intact, though a spreadsheet program may need UTF-8 chosen by hand on import.

Because it is line-oriented plain text, ordinary tools do the rest. Search it for a phrase to find the moment viewers reacted; split each line on the first closing bracket and then on the first colon after it to get timestamp, name and message as three spreadsheet columns; or count lines per minute to locate the loudest stretch before you start editing. The chat replay is read directly from Twitch's public GraphQL endpoint in your browser, 100 messages per request, with a running message count shown while it pages through, so the proxy that relays video segments is not involved at all.

The Chat (.txt) button stays hidden until you analyse a twitch.tv/videos/… link, which means it is offered for past broadcasts and highlights but not for a clip or a live channel. If that video has no chat replay stored, the export stops with the message "No chat available for this VOD" instead of writing an empty file. When the same VOD has chapter markers, the separate chapter export uses an identical [offset] prefix, so the two text files line up: take the segment boundaries from the chapter file, then pull the matching stretch of conversation out of the chat file by timestamp without touching the video.