YouTube Transcripts in Non-English Languages — What Works

Extracting transcripts from non-English YouTube videos works well with INDXR — but the free caption route and AI transcription give you different things, and it's worth knowing which to reach for.

What Caption Extraction Gives You for Non-English Videos

YouTube's auto-caption system generates captions in the video's original language. Arabic videos get Arabic captions. Spanish videos get Spanish captions. The catch that trips up most transcript tools is at the delivery level: when you fetch captions through YouTube's timedtext API, its CDN can force the output through an English translation layer with a tlang=en parameter — so you ask for Arabic and get English. The language field still reads "ar", but the text is the English translation.

INDXR avoids this. Instead of the translatable track, it anchors to the video's native track — the one YouTube marks as the original (its -orig track) — which never carries a tlang= translation. So for a non-English video that has auto-captions, caption extraction returns the original language, not an English translation. That's a deliberate choice in how INDXR requests captions, and it's where it differs from tools that take whatever the timedtext API hands back.

Where caption extraction still can't help is when a video has no captions at all, or when you want punctuation and clean sentences that auto-captions don't provide. For those, AI Transcription is the route.

What AI Transcription Gives You

AI Transcription downloads the video audio and runs it through AssemblyAI's speech recognition models directly — bypassing YouTube's caption system entirely.

For Arabic, Spanish, Portuguese, Turkish, Indonesian, and many other languages, INDXR transcribes the audio in its original language. We automatically pick the best model for the detected language — our highest-quality model, AssemblyAI Universal-3.5 Pro, natively covers 18 languages (including Arabic), and a broader model covers up to 99 languages for the rest.

The output is the actual spoken language, correctly transcribed, with punctuation. Here's a real example from an Arabic lecture video (Dr. Tariq Al-Suwaidan, 28.5 minutes):

{
  "extraction_method": "assemblyai",
  "language": "ar",
  "segments": [
    {
      "text": "كثير من الناس يُخصّصون كل جهدهم ووقتهم فقط للبحث عن المال",
      "start_time": 35.2,
      "end_time": 42.1
    }
  ]
}

Correct Arabic text. Correct timestamps. The same structure as any English transcript.

When to Use Each Approach

For RAG pipelines specifically: if you're building a knowledge base in Arabic, Turkish, or Indonesian, AI Transcription is the only reliable route to original-language chunks.

Cost

AI Transcription: 1 credit per minute, minimum 1 credit.

A 28-minute Arabic lecture: 28 credits. At Plus pricing, that's €0.70.

For Arabic specifically, our highest-quality model handles Modern Standard Arabic and many dialects reliably. When a language falls outside the highest-quality model, INDXR automatically uses a broad-coverage model spanning 99 languages — so you always get original-language text.

For the full JSON export schema, see YouTube Transcript JSON Export. For audio file uploads, see Audio Upload. For credit packages, see the pricing page.