Transcribe a video link on my Mac
Transcribe a link: the speech at a web address (a YouTube video, a podcast episode, any page that plays sound), read from the page's own subtitles or heard on the person's own Mac when it has none
Runs on the person's Mac: reads the page at url with the Mac's link reader (computer/hear-a-link.json fetchers, such as yt-dlp). When the page has subtitles in the languages asked for, they are the transcript; otherwise its sound is fetched and heard on the Mac with its own speech model (the same hearing as tool:hear-a-recording-on-this-mac), and the sound is kept in the person's files under /Heard links with the transcript on it, so search finds it. Answers { url, title, duration_seconds, text, segments, language, source: captions or heard, file }. A link with no speech fails with the reason, never an empty answer.
What it asks for
- url (Needed, words): The link to the video or recording
- languages (Optional, words): Caption languages to look for, such as "en.*,en" or "es"; the Mac's own list when left out
- keep_sound (Optional, yes or no): false to not keep the heard sound in the person's files
Use it from your AI
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to transcribe a link: the speech at a web address (a YouTube video, a podcast episode, any page that plays sound), read from the page's own subtitles or heard on the person's own Mac when it has none”
What your AI calls: run_tool {"tool_id":"tool:hear-a-link-on-this-mac"}