YouTube Transcript API

What the API does

The Transcript API is for developers who want the bulk YouTube transcript capability of YouTube Video Transcript inside their own product or workflow. Paste a YouTube URL into your code, get back a clean JSON payload with every caption line, timestamps, and video metadata, ready to stream into a vector store, a database, or an LLM prompt.

It is designed around the three shapes of request that actually matter. A single video URL hits a sync endpoint and returns the transcript in one HTTP call. A playlist or channel URL goes through the bulk job flow: submit the job, poll the status endpoint, and pull a ZIP when it completes. The API handles enumeration and caption-track selection so you do not touch the YouTube Data API, rotate proxies, or reason about caption-track variants.

Authentication uses a bearer token tied to your account API key. The single-video endpoint returns either JSON (default, with caption lines and timestamps for direct ingestion into vector stores, databases, or LLM prompts) or a formatted file when you pass ?format=txt|srt|csv|docx|markdown. Bulk-job exports return a ZIP of per-video files. The full response shapes are documented at /docs. The underlying transcript engine is the same one that powers YouTube Video Transcript exports, so it inherits the same handling of YouTube caption-track edge cases.

Pricing

The API shares the same credit pool as the dashboard, so a transcript fetched through the API counts the same as one fetched through the UI. That keeps plans legible and avoids the classic trap of having to pick between two separate SKUs. Free covers 10 lifetime transcripts; paid plans start at $9/mo for 1,000 transcripts. See pricing for the full tier breakdown.

Use cases

The API is built for three families of use case. If you are doing one of these, expect a near-zero integration cost. The endpoint and response shape are designed around them.

Apps and SaaS products

If you are shipping a feature that summarizes, searches, or annotates YouTube videos, the transcript is the hardest part to get right. The API gives you reliable captions with timestamps so you can ship video summarization, chaptering, quote extraction, or a citation feature without building your own caption pipeline. Think note-taking apps, research tools, learning platforms, and content-repurposing SaaS.

AI agents and LLM pipelines

Agents that read video content as part of a larger task (a research agent, a knowledge-base builder, a retrieval-augmented-generation pipeline) need captions in a shape that chunks well for embeddings. The API returns per-line JSON with timestamps so chunking by token or by time window is a one-liner. Pair it with a summary prompt and you have a production-grade YouTube tool for your agent stack.

Automation and data workflows

For researchers, journalists, and data teams who already run scheduled jobs, the API is a drop-in replacement for hand-rolled yt-dlp scripts. Call it from a cron job, an Airflow DAG, a Temporal workflow, or a Zapier step. Every endpoint has a documented response shape, so the pipeline code never needs to handle the weird-case YouTube URL variants.

Comparison vs RapidAPI providers

The existing market for YouTube transcript APIs lives mostly on RapidAPI, where a dozen providers resell the same underlying caption scrapers with inconsistent shapes and hard-to-predict pricing. A handful of them are decent, most of them are not. Here is the honest picture.

• Pricing clarity. RapidAPI pricing mixes request limits, subscription tiers, and burst allowances in ways that punish you for growing. We charge a flat credit price per transcript, the same credit a UI download costs.
• Bulk support. Most RapidAPI providers charge per video and offer no channel or playlist endpoint. We wrap enumeration in an async bulk job so one POST submits the whole channel and one GET pulls the resulting ZIP.
• Response stability. RapidAPI transcript APIs break regularly when YouTube changes its caption endpoints. Our backend fetches captions directly with an independent fallback path, so a single upstream change does not take the whole service down. It has been running in production against real customer traffic.
• Direct support. RapidAPI support is the provider's marketplace inbox. Our support is a real email from the team that built it.

What is shipping today

The live API has two shapes. Single videos hit a sync endpoint and return either JSON or a formatted file (TXT/SRT/JSON/CSV/DOCX/Markdown). Playlists and channels go through the async job flow: POST to submit, GET to poll status and per-video results, GET to pull a ZIP when the job finishes. Full method, parameter, and error reference lives at /docs. For how this API stacks up against the other developer options, see our comparison of the best YouTube transcript APIs.

The same backend serves the MCP connector: install it once in Claude.ai, Claude Desktop, ChatGPT, or Grok and your chat client can fetch any YouTube transcript on demand without copy-pasting. Same account, same credit pool, same backend. Claude.ai, Claude Desktop, and ChatGPT install via OAuth consent; Grok takes your yvt_live_ API key pasted into its Custom Connector dialog.