Quotes
Earlier project
I wanted a way to find a line from a television episode and jump to the matching clip. Quotes grew out of that idea, using subtitle data to connect searchable dialogue to timestamps in the original video.
Building the quote library
I used FFmpeg to extract subtitles from episode files, parsed the subtitle data, and stored dialogue alongside episode information and start and end timestamps. That turned the captions into a searchable database.
From a quote to a clip
The clip-delivery implementation used a Go backend and FFmpeg to trim the matching segment on demand and convert it into a format the browser could play. A temporary cache avoided repeating the same processing when a browser made multiple requests for a clip.
The web application connected a selected quote to the episode and timestamps needed by the video backend. Keeping the text and timing information together made that link possible.
Project history
This was a later iteration of my television-quotes projects, following an earlier Office Quotes site. The focus shifted from browsing transcripts to connecting a search result directly to a playable moment.