YouTube Transcript Extractor: Batch, Timestamps, RAG Chunks
Pricing
from $2.40 / 1,000 transcript extracteds
Batch YouTube transcripts from video or playlist URLs: clean text, timestamped segments and RAG-ready chunks with deep links. No API key.
YouTube video URLs, 11-character video IDs, or playlist URLs (first 100 videos of a playlist). Short youtu.be and /shorts/ links work.
Language code such as en, de, es. Falls back to the first available track (auto-generated tracks included).
Add a segments array with start time and duration for each caption line.
If above 0, output a chunks array of about this many characters (never more than 1.4x), cut at natural pauses and sentence ends. Auto-generated captions are split by word-level timing gaps, because they have no punctuation. Each chunk has its start time, a deep link, and textWithContext (video title and channel prepended) for embedding. Keep it at or below ~1500 for 512-token embedding models. 0 disables.
Write each chunk as its own row (videoId, title, channel, chunkIndex, startSeconds, link, text, textWithContext) to a separate dataset; the main output keeps one item per video with chunksDatasetId instead of the full text. Recommended for AI agents and MCP clients: rows can be paged with limit/offset and projected with fields, so the agent never receives a whole 45k-token transcript in one tool result. Price stays per video. Requires chunkChars above 0.
