YouTube videos often have captions in more than one language — and you can pull the transcript in the one you actually need. This guide explains how caption languages work, how to choose one, and what to do when the language you want isn't there.

Three kinds of caption languages

Not all captions are equal. A video can have:

Human-authored captions — written by the creator (or their team). Highest quality, but only in the languages they bothered to add.
Auto-generated (ASR) captions — YouTube's speech recognition, usually in the video's spoken language. Good, with the occasional error.
Auto-translated captions — YouTube machine-translates an existing track into other languages on the fly. Widely available, but it's machine translation, so treat it as a draft.

So a Spanish-language video might offer authored Spanish, plus auto-translated English, French, German, and dozens more.

See what's available first

Before you extract, it's worth knowing which languages a video actually has. With Scribefy you can check the available caption tracks — each one tagged as authored or auto-generated — so you're not guessing. Via the API that's the free /api/info endpoint, and in Claude it's the get_video_metadata tool, both of which list every track with its language code. (More in the developer guide.)

Get the transcript in a specific language

On the web app, paste the URL and pick the language you want from the available tracks. Programmatically, pass a BCP-47 language code in the lang parameter:

curl -X POST https://api.scribefy.app/api/extract \
  -H "Authorization: Bearer sk_live_…" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://www.youtube.com/watch?v=…", "lang": "es"}'

Common codes: en (English), es (Spanish), fr (French), de (German), pt (Portuguese), hi (Hindi), ja (Japanese), zh-Hans (Simplified Chinese). Omit lang and you'll get English by default.

When your language isn't available

If you request a language the video doesn't have, the API returns a LANG_UNAVAILABLE error rather than guessing. Two good options:

Pick an available track. Check the metadata first and choose the closest one — often the authored original or YouTube's auto-translation.
Extract once, translate with AI. Pull the transcript in whatever language is available, then ask Claude or ChatGPT to translate it. Because the transcript is timestamped, the translation keeps its structure — and an LLM translation is often cleaner than YouTube's auto-translate. See summarizing (and translating) a video with Claude.

Why this beats copy-pasting

YouTube's own transcript panel shows one language at a time and strips formatting when you copy it. Pulling a specific language as a clean, timestamped transcript means you can drop it straight into notes, subtitles, or an AI prompt — in the language your audience reads.

Frequently asked questions

Can I get a transcript translated into a language the video doesn't offer?

Not directly — extraction returns YouTube's existing tracks. But you can extract an available language and translate it with an AI assistant in the same conversation; for timestamped source text, the result stays well-structured.

How accurate are auto-translated captions?

They're machine translation of (often already auto-generated) captions, so quality varies. For anything important, prefer an authored track if one exists, or translate a clean transcript with an LLM yourself.

Does a different language cost more?

No — pricing is by video length, not language. A fresh extraction is 1–8 credits and cached transcripts are free, regardless of language. See pricing.

What about videos with no captions at all?

If a video has no human or auto-generated captions in any language, there's nothing to extract. See transcripts without captions.

Need a transcript in a specific language? Paste the URL and pick your track — or pass lang via the API.