# Changelog of YouTube Transcript, Subtitles & Comments Scraper (`aaronoptimus/youtube-transcripts-scraper`) Actor

- **URL**: https://apify.com/aaronoptimus/youtube-transcripts-scraper/changelog.md
- **Full Actor documentation**: https://apify.com/aaronoptimus/youtube-transcripts-scraper.md

## Changelog

### Price cut — October 2026

- Transcripts now cost $2 per 1,000 (from $1.60 on higher Apify plans), down from $3 (from $2.40). Comments and the run start fee are unchanged.

### Search, AI chunks & monitoring — October 2026

- **YouTube search**: `searchQueries` (or `youtube.com/results?search_query=…` URLs) with *Max videos per search*, YouTube's relevance / popularity sort, upload-date filter and Subtitles/CC filter. Searching is free.
- **LLM / RAG output**: `transcript.chunks` by seconds and/or words (never across chapters, each with a link to its moment and its chapter), `transcript.paragraphText`, and *Lean items* without SRT / VTT / segments. Rows output gets `transcript_chunk` rows and a `timestampUrl` column.
- **Ordered language fallback**: `languages` list (the single `language` field still works) and *If none of these languages exists*: original, translate, or skip (not charged).
- **Channel monitoring**: *Published after* (date or age), *Include Shorts*, *Include live streams*, and *Skip videos already scraped* (IDs kept in a named key-value store, off by default).
- Channel and playlist listing reads YouTube's new view models (lockup / Shorts lockup), so channels and playlists expand again; live and scheduled videos are left out of listings.
- New prices: $3 per 1,000 transcripts (from $2.40 on higher Apify plans), comments $0.50 per 1,000 (from $0.40), run start $0.00005.
- Free Apify plan: up to 100 videos per run (was 10).

### 1.0 — October 2026

First public release.

- Transcripts of YouTube videos as plain text, SRT, WebVTT and timed segments, with the list of available caption languages.
- Caption language selection (uploaded captions before auto-generated ones, exact code before base language) and translation through YouTube's own `tlang` translation.
- Video info from the iOS app client first (then Android, then web; YouTube's web client needs a proof-of-origin token since October 2026). Three transcript sources with automatic fallback: the watch-page transcript panel (web-client info only) → the caption-track download → the Android player's caption tracks (no proof-of-origin token needed).
- Video metadata (title, channel, publish date, duration, views, likes, description, tags, category, thumbnails) and **chapters** (YouTube markers, else timestamps in the description).
- **Comments** with replies, top / newest order, author, likes, reply count, pinned / hearted / creator flags, approximate dates.
- **Channel and playlist URLs** (@handles, /channel/, /c/, /user/, playlists, bare IDs) expanded to their videos, newest first, capped per source, free of charge.
- **Flat rows** output format for CSV / Excel / Google Sheets.
- Reliability: one InnerTube session per worker with a consistent fingerprint, session/IP rotation on 403 / 429 / 5xx, sign-in walls and parser failures, capped backoff, per-video time limit, scheduled rotation, IPs that YouTube accepts shared between workers (at most two per IP), a temporary per-video residential fallback, parallel sessions, resume after restarts.
- Pay only for delivered transcripts and comments; videos without captions, private, removed, age-restricted, members-only and live videos are delivered free with a reason; videos that stay blocked are listed as errors and not charged; *Maximum cost per run* is respected across both event types and all parallel sessions.
- Crash- and migration-safe billing: progress is saved before every charge, so a restarted or migrated run never charges the same video twice.
- Free Apify plan: runs are limited to 10 videos and 50 comments per video and finish normally.