TranscriptFetch

About TranscriptFetch

TranscriptFetch is an API that converts videos from YouTube, TikTok, and Instagram into structured, timestamped transcripts. It resolves the fastest available source for a given video: checking a cache, reading published captions when they exist, and transcribing audio when captions are unavailable, returning a consistent JSON schema of segments with start times, language, and duration in every case. It also supports resolving a channel or playlist into a paginated list of videos, with polling via since_video_id that does not consume credits when no new videos are found.

The product targets developers and AI builders working with spoken-word content, offering a REST API and a hosted MCP server (with a prebuilt n8n node) so that clients such as Claude, ChatGPT, Cursor, and VS Code can fetch transcripts directly. Listed use cases include continuous monitoring of channels or playlists for new uploads, building research datasets from interviews, reviews, demonstrations, or lectures, and giving agents access to primary spoken sources with exact quotes and timestamps.

Billing is credit-based: caption-sourced transcripts cost 1 credit, audio transcription is billed per started minute, and failed requests are not charged. The page reports over 1,700,000 transcripts fetched, 99.82% API uptime, and an average response time of 2.9 seconds over the last 30 days.

Key features

• Fetches timestamped transcripts from YouTube, TikTok, and Instagram video URLs
• Transcribes audio automatically when no caption track exists
• Lists videos from a channel or playlist for polling new uploads
• Continuous monitoring of channels/playlists with searchable archive of new uploads
• Hosted MCP server plus prebuilt n8n node for agent access (Claude, ChatGPT, Cursor, VS Code)

Frequently asked questions