YouTube Transcript in the Wrong Language? It's Probably Auto-Dubbing
An English video whose transcript comes back in Arabic or Hindi has almost always been auto-dubbed. Why it happens, how to spot the original language, and how to get what was actually said.
Short answer
If a YouTube transcript comes back in a language nobody speaks in the video, the video has almost certainly been auto-dubbed. YouTube now creates AI voice-overs in up to 20 languages and gives each one its own caption track. The transcript you want is the one in the video's original language — or an AI transcript of the original audio.
You paste a YouTube link to an English talk, and the transcript comes back in Arabic. Or Hindi, or Ukrainian. The speaker never said a word in that language. Nothing is broken on your end: the video has been auto-dubbed, and the transcript you got is of a dub.
What auto-dubbing does to a video
YouTube can now dub a video automatically with AI voices. In February 2026 it made auto-dubbing broadly available and extended it to 27 languages; from English, it dubs into 20 of them. A dubbed video carries extra audio tracks — one per language — and YouTube generates captions for each dub just as it does for the original.
On one English video we checked this month, YouTube listed 20 caption tracks: Arabic, Bengali, Dutch, French, German, Hebrew, Hindi, Indonesian, Italian, Japanese, Korean, Malayalam, Polish, Portuguese, Punjabi, Russian, Spanish, Tamil, Telugu and Ukrainian. Every one of them is the transcript of a dub.
Why you got the wrong language
Two things combine.
The tracks are listed alphabetically. YouTube orders caption tracks by language name, so on a dubbed video the first track is usually Arabic. Any tool that takes "the first caption track" when it can't find English hands you the Arabic dub.
The original captions can arrive late. On a fresh upload, the dubs' captions can appear before the captions in the language the creator actually spoke. On that English video, YouTube offered the 20 dub tracks and no English one for at least a day; asking for English captions directly returned nothing. The English track appeared the following day.
So for a while the only captions on offer were translations, and the "first" of them was Arabic.
How to tell what language a video is really in
On YouTube, open the video's settings (the gear icon) and choose Audio track. On an auto-dubbed video, one track is labelled original — that's the language the creator spoke. If there is no Audio track option, the video has one track and it is the original.
In the captions menu, a track ending in "(auto-generated)" was made by speech recognition. On a dubbed video several tracks carry that label, because the dubs are auto-generated too.
Getting the transcript of what was actually said
Paste the link into SubGrab. Since October 9, 2026:
- English captions first. If the video has captions in English, you get them, as before.
- Otherwise, the language the video is spoken in. SubGrab reads YouTube's "original" audio label and picks the captions in that language — the speaker's own words, not a translation.
- No captions in the original language yet? Then there is nothing faithful to extract, so instead of a dub's captions you are offered an AI transcript of the original audio (one credit per started hour, shown before you confirm). It downloads the original audio track specifically, not a dub.
The language menu on the transcript page lists every track YouTube publishes for the video, by name — "Arabic (auto-generated)" rather than a bare "ar" — so choosing a dub is a deliberate choice, not an accident.
Which kind of caption track is which
| Track | What it is | Faithful to the speaker? |
|---|---|---|
| Original-language, auto-generated | Speech recognition on the creator's audio | Yes |
| Uploaded by the creator | Captions the creator wrote or edited | Yes |
| A dub's track, auto-generated | Speech recognition on an AI voice-over | No — it is a translation |
| Auto-translated | Machine translation of another track | No |
A translated transcript is perfectly good for understanding a video in your own language. It is not a record of the exact words, which matters for quotes, research, subtitles and anything you will cite. For translated subtitles done properly, see transcribing in 90+ languages and the difference between captions, transcripts and subtitles.
If the video has no captions at all
Plenty of videos — auto-dubbed or not — publish no captions. SubGrab then transcribes the audio with AI, in the language it hears. The full walkthrough is in how to get a YouTube transcript.
Frequently asked
Why is my YouTube transcript in Arabic when the video is in English?
Because YouTube auto-dubbed the video and the tool you used picked a dub's captions. Each dub gets its own caption track, and YouTube lists the tracks alphabetically by language name, so Arabic is usually first. A transcript of the Arabic dub is a translation, not what the speaker said.
How do I find out which language a YouTube video was originally recorded in?
Open the video's settings (the gear icon) and choose Audio track. On an auto-dubbed video, one track is labelled original — that is the language the creator spoke. A video with no Audio track option has only one track, which is the original.
Can I still get the transcript in another language?
Yes. SubGrab's language menu on a transcript lists every caption track YouTube publishes for the video, including the dubs, with their names. Picking one gives you that language's track; it is a translation of the original, which is fine for reading but not a record of the exact words.
What does SubGrab do with an auto-dubbed video now?
It gives you English captions when the video has them, otherwise the captions in the language the video is spoken in. If YouTube only has captions for the dubs — which happens for a day or so after some uploads — it offers an AI transcript of the original audio instead of a dub's captions.