Skip to main content

YouTube subtitles API with multi-language support

Download YouTube subtitles in the language you need. One endpoint for transcript text, another to list every available language on a video.

  • Language parameter with automatic fallback
  • Per-segment timestamps for subtitle rendering
  • MCP endpoint for AI assistant integrations

Subtitles and captions are the same underlying data on YouTube: timed text synchronized to playback. A subtitles API lets you request those lines in a chosen language, with fallbacks when the preferred locale is missing. FreeTranscriptAPI exposes language selection on every transcript request and a dedicated languages listing so you do not hard-code assumptions per video.

Choosing a language

Pass lang on /transcript with an ISO-style code such as en, de, or es. When the exact track is unavailable, the service applies documented fallback behavior instead of returning an empty body without explanation.

For UI that lets users pick a language, populate a dropdown from /languages at runtime. Store both the requested code and the resolved language from the transcript response so support can reproduce issues when creators add new tracks later.

Rendering subtitles in your player

Each JSON segment includes start time and duration in seconds. Multiply or format those values into SRT, WebVTT, or your player's native cue format. Because durations are supplied per line, you can detect overlapping cues or zero-length rows in QA before shipping to end users.

If you only display text under the video without timing, format=text still respects language selection while skipping segment metadata.

Translation workflows

The API returns captions YouTube already hosts. It does not machine-translate on the fly. Typical workflows fetch the source language, send text to your translation provider, and store derived subtitles separately. Timestamps from the original track usually remain valid when line breaks are preserved.

When creators upload manual translations on YouTube, prefer those tracks via the language code YouTube assigns rather than re-translating auto-generated text, which reduces error compounding.

Batch and multi-video jobs

Playlist or channel crawlers should respect rate limits: stagger requests, backoff on 429, and use an API key so credits accrue to your account instead of shared anonymous quotas. Pass the same lang across a batch for consistent corpora, or vary it per locale when building multilingual datasets.

Example request

curl "https://api.freetranscriptapi.com/v1/transcript?video_url=https://www.youtube.com/watch?v=dQw4w9WgXcQ"

MCP and assistant integrations

AI assistants can call FreeTranscriptAPI through the hosted MCP endpoint documented in the API reference. That lets tools in Cursor or other MCP clients fetch subtitles during a coding session without embedding API keys in local config beyond your normal secret store.

Related reading

Compare hosted APIs with npm scrapers in the npm vs API article, or see how free tiers stack up in the free API comparison.

Anonymous callers get 20 requests per hour per IP without an API key.