YouTube Transcript Fetcher (Apify)

A fetcher skill that pulls the subtitle/transcript text of YouTube videos in bulk. It provides a standard interface to the Apify streamers/youtube-scraper actor. Because the YouTube Data API v3 does not return transcripts and yt-dlp gets rate-limited (429), this actor is in practice the only reliable source.

What does it do: It takes a list of video IDs or full URLs (up to 50 videos at a time), downloads the auto-generated or manual subtitles for each video, cleans and dedupes the timestamps, and returns plain transcript text. It works multilingually (en, tr, es, ...). It keeps no quota file; pay-per-use (~$0.001-0.01/video).

When is it used: When you want to extract the spoken text of the top 10 videos in a creator analysis, summarize podcast/long-form content, study hook and retention science from a transcript, get the auto-generated subtitles of a Turkish YouTuber, or request raw SRT for multi-speaker analysis. For metadata (views, channel statistics), use youtube-data-fetcher, not this — this skill is for TRANSCRIPT only.

Output: One JSON line (jsonl) per video — video_id, url, title, duration, view_count, channel, language, has_subs, word_count, and the cleaned transcript. Videos with subtitles disabled are reported with has_subs: false, and the batch does not error out. Credentials are read from APIFY_API_TOKEN in ~/.claude/.env.