transcriptyt

transcriptyt

Python client for the TranscriptYT YouTube Transcript API. Get any public video's transcript as JSON, text, SRT, VTT, CSV or Markdown, in 150+ languages, with AI transcription when a video has no captions. Requests run through rotating proxies, so they keep working from cloud servers where YouTube blocks direct scraping.

pip install transcriptyt

Get a free API key (100 credits) at transcript-yt.com/dashboard/keys.

Usage

from transcriptyt import TranscriptYT

client = TranscriptYT()  # reads TRANSCRIPTYT_API_KEY, or pass api_key="ts_live_..."

t = client.transcript("https://www.youtube.com/watch?v=dQw4w9WgXcQ")
print(t["title"], t["language"])
print(t["text"])
for seg in t["segments"][:3]:
    print(seg["timestamp"], seg["text"])

# Other formats return a string
srt = client.transcript("dQw4w9WgXcQ", format="srt")

# Translate, or pick a language
client.transcript("dQw4w9WgXcQ", translate_to="es")
client.transcript("dQw4w9WgXcQ", language="de")

# Up to 50 videos per call, or a whole playlist/channel
batch = client.transcripts(["dQw4w9WgXcQ", "jNQXAC9IVRw"], format="text")
batch = client.transcripts(playlist="https://www.youtube.com/@mkbhd", limit=20)
for r in batch["results"]:
    print(r["url"], r["ok"])

# Free endpoints
client.search("python tutorial", limit=10)
client.playlist("https://www.youtube.com/playlist?list=...")
client.languages("dQw4w9WgXcQ")
client.usage()

# AI summary, key points and timestamped notes
client.summary("dQw4w9WgXcQ")

Async

from transcriptyt import AsyncTranscriptYT

async with AsyncTranscriptYT() as client:
    t = await client.transcript("dQw4w9WgXcQ")

Errors

Every non-2xx response raises TranscriptYTError with .code (e.g. NO_CAPTIONS, VIDEO_UNAVAILABLE, RATE_LIMITED, OUT_OF_CREDITS), .status and .message. Rate-limit and capacity responses (429/503) are retried automatically (max_retries=2).

from transcriptyt import TranscriptYTError

try:
    client.transcript("dQw4w9WgXcQ", mode="captions")
except TranscriptYTError as e:
    print(e.code, e.message)

Transcript options

Argument	Description
language	BCP-47 code of the track to fetch
translate_to	Machine-translate to this language (150+), keeping timing
format	json (default), text, srt, vtt, csv, md
include_segments	False returns only the full text (JSON)
timestamps	Prefix lines with [HH:MM:SS] in text/md
paragraphs, chunk_size	Merge segments into paragraphs of up to chunk_size characters
caption_type	any, manual, auto
original_language	Prefer the spoken language over English
mode	auto (captions, then AI), captions (never AI), ai
start_time, end_time	Only part of the video, in seconds
search	Only lines containing this text

Full API reference: transcript-yt.com/docs. Using LangChain? See langchain-transcriptyt.

License

MIT