Auto-Dubbed YouTube Videos: How to Find the Original Transcript

Some YouTube videos now ship with auto-dubbed audio in several languages — the same video, with an AI-generated voice track standing in for the creator's own. If you need the creator's actual words rather than a dubbed rendition, you need to know which track is original, and YouTube doesn't always make that obvious from the video page alone.

How auto-dubbing differs from captions

Auto-dubbing replaces the audio track itself with a synthesized voice in another language; it's a different mechanism from captions, which overlay text on top of whatever audio is playing. A video can have auto-dubbed audio in several languages and still have its own separate set of caption tracks per language — is_original specifically marks the audio track, so it's the right signal to check when your concern is which spoken version you're transcribing, not which caption track you're reading.

Why this matters for a transcript

A transcript pulled from a dubbed audio track reflects whatever that dub says — which is a translation (and sometimes a loose one) of the original, not the original itself. If you're quoting someone, citing a statement, or doing any kind of research where exact wording matters, you want the original-audio track specifically, not just whichever language happens to match your own.

How to tell them apart

The languages endpoint marks exactly this. Each track returned for a video includes an is_original flag, set to true on the track that matches the creator's actual spoken audio:

curl https://scriptube.me/api/v1/videos/VIDEO_ID/languages \
  -H "X-API-Key: YOUR_API_KEY"

Everything else returned alongside it — other native tracks, auto-dubbed languages, translation targets — is not the original, regardless of how it's labeled on YouTube's own player.

A worked example

Say a video was originally uploaded in Portuguese, and YouTube has auto-dubbed it into English, Spanish, and French. Checking /languages for that video returns all four as available tracks, but only the Portuguese one comes back with is_original: true. If your goal is an accurate transcript of what the creator actually said, that's the one to request — not the English dub, even if English is the language you'd naturally reach for.

How to do it in Scriptube

1. Check /languages for the video before requesting a transcript, if you're not sure which track is original.
2. Find the entry with is_original: true and note its language code.
3. Request the transcript with language=xx using that code.

curl "https://scriptube.me/api/v1/transcripts/{batch_id}?language=pt" \
  -H "X-API-Key: YOUR_API_KEY"

From the dashboard, the language selector shows the same underlying tracks — if a video has auto-dubbed audio, picking the original-language option gets you the creator's actual words rather than a dub.

Most videos don't need this at all

Auto-dubbing is still the exception rather than the norm across YouTube — most videos have exactly one audio track, which is trivially the original. The is_original flag matters specifically for the subset of videos that have been auto-dubbed into multiple languages; checking /languages costs nothing, so it's a reasonable first step whenever you're not certain a video falls into that category.

Combining this with translation

Finding the original track and translating a transcript are two different things that can be used together. You might pull the original-audio transcript first (using is_original to confirm you have the right track), then separately request an auto-translated version of it into a language you actually read — rather than translating a dub, which would be a translation of a translation. See translating a YouTube transcript into any language for how the translation flags work once you've identified the right source track.

If you want the dub instead

There's nothing wrong with wanting a dubbed track — if you're building content for an audience that only reads one language and accuracy to the single word isn't the goal, a dub (or an auto-translated transcript) is often good enough and faster to work with. The point of is_original isn't to discourage using dubs, it's to make sure you know which one you're getting before you rely on it for something where the distinction matters.

Quick answers to common questions

Can a video have no track marked as original? No — every video has one actual source-audio track; is_original identifies it even when several dubbed tracks are also available.

Does checking /languages cost anything? No, it's a free, read-only check — only submitting or requesting an actual transcript uses a credit.

Is the original track always the highest-quality captions? Not necessarily — caption quality depends on whether it's auto-generated or creator-uploaded, which is a separate question from which audio track is original. See auto-generated vs. manual captions.

Related reading

For the full picture of how language selection, native tracks, and auto-translation fit together, see translating a YouTube transcript into any language. If you specifically want two languages side by side — say, the original and a dub, or the original and a translation — see bilingual transcripts, side by side. For native caption tracks without any dubbing involved, see downloading YouTube captions in multiple languages.

When this comes up most

This matters most for research, journalism, and any context where quoting someone precisely is the point — educational channels, news clips, and large creators are the ones most likely to have auto-dubbed versions in the first place, so it's worth a quick /languages check specifically on that kind of content rather than assuming the audio you're hearing is what the speaker actually said.

Check a video's available tracks and which one is original with an API key from your dashboard, or see the full languages endpoint in the API docs.