How do I get a YouTube transcript in a specific language?

Q: How do I get a YouTube transcript in a specific language?

Pass lang (for example en or es) to the YouTube transcript endpoint; it returns that track when the video has one, and the response carries a lang field you can check before storing the text. Each transcript costs $0.004, so check lang before asking again, and translate the segments yourself when the video only has another language.

Say you track 50 Spanish-language cooking channels for a US food brand, and your team reads English. Some of those videos carry an English caption track, most only have Spanish. You want English wherever YouTube has it and a clean signal for the rest, without paying twice per video.

Asking for a language

YouTube video transcript takes id (the 11-character video ID) and an optional lang with a language code such as en, es or ja. Treat lang as a preference, not a promise. Which tracks a video offers is up to the creator and YouTube, so the transcript you get back carries its own lang. Read it every time before you store the text under a language.

A pattern that pays once per video

import os, requests

API = "https://api.okdata.online/v1/youtube/videos/transcript/get"
H = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}

def transcript(video_id, want="en"):
    body = requests.get(API, headers=H, params={"id": video_id, "lang": want}, timeout=60).json()
    if body["code"] == "not_found":
        return None                      # no captions at all; not charged
    if body["code"] != "ok":
        raise RuntimeError(f"{body['code']}: {body['message']}")
    t = body["data"]["item"]
    t["needs_translation"] = t["lang"].split("-")[0] != want
    return t

t = transcript("VIDEO_ID", want="en")
if t and t["needs_translation"]:
    print("Got", t["lang"], "- translate", len(t["segments"]), "segments")

Translate segment by segment, not the whole text, and the start and end timings stay valid for quotes and subtitles.

Cost

Plan	Calls	Cost
50 videos, one language request each	50	$0.20
Same 50 videos, a second language asked for every video	100	$0.40
1,000 videos	1000	$4.00

Every returned transcript is a charged call, even when it came back in a different language than you asked for. That is why the code above asks once and translates locally instead of looping through codes.

Gotchas

Use the language part of the code when you compare (en-US and en are the same language for most work).
On videos with several caption tracks, the lang label can name the video's first listed track rather than the one you got. When the language decides what happens next, run a language detector on a few segments as a cross-check.
The response does not mark whether a track was uploaded or generated automatically.
Videos with no captions return not_found at no charge. There is no speech-to-text step.
If you come from youtube-transcript-api: its find_transcript expects a list of codes, and translated tracks fail for some videos. Here you send one code and compare.

Other routes

On your own channel, YouTube Studio and the official Data API can add or download caption tracks in any language you upload.
For a handful of videos, YouTube's player can show auto-translated captions to a human reader for free.
When quality matters, a professional translation of the transcript beats any automatic track.

More on batch transcript work is on the video transcript solution page.