TikTok & YouTube Transcript Scraper

Turn any TikTok or YouTube video into text. This Actor runs speech-to-text on the video's audio, so it returns a transcript even when the video has no captions or subtitles. Paste video URLs, get clean transcripts with per-phrase timestamps — or switch to a cheaper metadata-only mode when you just need titles, captions and durations. Built and maintained by Painted Labs.

What does this Actor do?

• 🎙️ Speech-to-text transcription — runs on the actual audio. Works on videos that have no captions at all.
• ⏱️ Per-phrase timestamps — each transcript comes with { start, end, text } segments, ready for subtitles or clip search.
• 🌐 Automatic language detection — the spoken language is detected per video and returned in the output. No configuration needed.
• 📋 Metadata-only mode — flip one switch to skip transcription and fetch just the title, description/caption and duration, at a much lower price.
• 🔗 TikTok and YouTube in one run — tiktok.com, youtube.com and youtu.be links, mixed freely.
• 👤 Whole profiles, channels & playlists — paste a TikTok profile or YouTube channel/playlist URL and its newest videos are pulled in automatically, up to a per-source limit you set.
• 🔍 YouTube keyword search — turn a search query straight into transcripts of the top matching videos.
• 🎛️ Date & popularity filters — when pulling from profiles, channels or search, keep only videos from a date range or above a view count, newest-first or most-viewed-first.

What data you get

Field	Description
url	The video URL you submitted
success	Whether the video was processed
transcript	The spoken words, as text (empty in metadata-only mode or when there is no speech)
segments	Timestamped { start, end, text } phrases (transcription mode)
language	Detected spoken language
durationSeconds	Video length in seconds
title	Video title
description	Video caption / description
viewCount	View count at scrape time (videos discovered via profiles/channels/search)
postedAt	Post date (videos discovered via profiles/channels/search)
jobId	Internal id for the processed video

How to use it

1. Open the Actor and press Try for free.
2. Paste your TikTok / YouTube links into Video URLs, one per line — and/or add profile, channel or playlist URLs (or a YouTube search) to pull whole feeds.
3. Leave Transcribe audio on for full transcripts, or switch it off for metadata only.
4. Press Start. Results stream into the dataset as each video finishes.
5. Export the results as JSON, CSV, or Excel from the Storage tab.

Input

Field	Type	Default	Description
videoUrls	array	—	TikTok or YouTube video links, one per line.
profileUrls	array	—	TikTok usernames (@user) or profile URLs, YouTube channel or playlist URLs — each expands into its newest videos.
searchQueries	array	—	YouTube keyword searches — each expands into the top matching videos.
maxVideosPerSource	integer	30	Videos taken from each profile, channel, playlist, or search query (1–1,000).
postedAfter / postedBefore	date	—	Only include discovered videos posted in this date range.
minViews	integer	—	Only include discovered videos with at least this many views.
sourceSorting	string	newest	Take each source's videos newest-first or most-viewed-first (mostViewed).
transcribe	boolean	true	On: speech-to-text with timestamps. Off: metadata only (title, caption, duration).
includeSegments	boolean	true	Include timestamped { start, end, text } segments with each transcript.
subtitleFormat	string	none	Also render each transcript as a subtitle file: srt or vtt. Free.
translateTo	string	none	Add a translation of each transcript in the translation field ($0.01/video, only charged when delivered). Requires transcribe: true.
maxConcurrency	integer	4	Number of videos processed in parallel (1–8).
perVideoTimeoutSecs	integer	900	Give up on a single video after this many seconds (60–3600).

At least one of videoUrls, profileUrls, or searchQueries is required. The date/views filters and sorting apply to videos discovered through profiles, channels, playlists and search — direct videoUrls are always processed as given. When a platform's listing doesn't expose a video's post date or view count, an active filter skips that video.

{
    "videoUrls": [
        "https://www.tiktok.com/@username/video/7301234567890123456",
        "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
    ],
    "profileUrls": ["https://www.tiktok.com/@username"],
    "searchQueries": ["keyword to look up on youtube"],
    "maxVideosPerSource": 30,
    "transcribe": true
}

Output

One dataset item per input URL. A transcription result:

{
    "url": "https://www.tiktok.com/@username/video/7301234567890123456",
    "success": true,
    "transcript": "Okay so in this video I'm going to walk you through the whole process step by step, it's easier than it looks and it takes about ten minutes...",
    "language": "en",
    "durationSeconds": 47.3,
    "title": "How to do it in 10 minutes",
    "description": "Full walkthrough in the video #howto #tutorial",
    "segments": [
        { "start": 0.0, "end": 3.4, "text": "Okay so in this video I'm going to walk you through the whole process" },
        { "start": 3.4, "end": 6.9, "text": "step by step, it's easier than it looks" }
    ],
    "jobId": "e9b1d2c4-5f6a-4b7c-8d9e-0f1a2b3c4d5e"
}

A metadata-only result (transcribe: false):

{
    "url": "https://www.tiktok.com/@username/video/7301234567890123456",
    "success": true,
    "title": "How to do it in 10 minutes",
    "description": "Full walkthrough in the video #howto #tutorial",
    "durationSeconds": 47.3,
    "jobId": "meta_766d7d3ea24fe0f309e6ff6e"
}

Special cases:

• No speech (music-only or photo/slideshow posts): success: true, noSpeech: true, transcript: "" — the caption is returned instead, and no transcription fee is charged. This is only reported when the audio was actually processed; an empty result that can't be verified comes back as a failure (EMPTY_RESULT_UNVERIFIED), not as "no speech".
• Failure (unavailable video, over the length limit, timeout, or unsupported URL): { "url": ..., "success": false, "error": "..." } — not charged.

Pricing

Event	Price	When
Result	$0.0018 ($1.80 / 1,000)	Every successfully processed video.
Transcription	+$0.018 per started minute	Only when audio is actually transcribed, once per started minute of video.
Translation	+$0.01 ($10 / 1,000)	Only when a requested translation is actually delivered.

A video under a minute — most short-form content — is one started minute: $0.0198 all-in, about $19.80 per 1,000 videos. A 3m20s video counts as 4 minutes ($0.0738 all-in). Metadata-only runs cost $1.80 per 1,000 videos. Failed videos, timeouts, unsupported URLs and no-speech videos are never charged the transcription fee.

Everything else is free: pulling videos from profiles, channels, playlists and search queries costs nothing (discovered videos are billed exactly like directly submitted URLs), and so are the date/popularity filters, sorting, timestamped segments, and SRT/VTT subtitle files. A failed translation is never charged.

Free Apify plan: limited to 5 videos and 1 transcription per month, without add-ons (translation, subtitle files). Any paid Apify plan unlocks everything.

What you can build with it

• 🤖 LLM & RAG pipelines — feed clean, timestamped speech into your models.
• 📈 Content research — analyse what creators actually say, not just what they caption.
• 🎬 Subtitles & clips — per-phrase segments drop straight into SRT/VTT and clip-search tools.
• 🔎 Search & indexing — make video audio searchable as text.

API usage

JavaScript (apify-client)

import { ApifyClient } from "apify-client";

const client = new ApifyClient({ token: "YOUR_APIFY_TOKEN" });

const run = await client.actor("paintedwinter/tiktok-youtube-transcript").call({
    videoUrls: [
        "https://www.tiktok.com/@username/video/7301234567890123456",
        "https://youtu.be/dQw4w9WgXcQ",
    ],
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const item of items) console.log(item.url, "->", item.transcript);

Python (apify-client)

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")

run = client.actor("paintedwinter/tiktok-youtube-transcript").call(run_input={
    "videoUrls": ["https://www.tiktok.com/@username/video/7301234567890123456"],
})

for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["url"], "->", item["transcript"])

curl (synchronous, small batches)

curl -X POST "https://api.apify.com/v2/acts/paintedwinter~tiktok-youtube-transcript/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" \
    -H "Content-Type: application/json" \
    -d '{ "videoUrls": ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"] }'

run-sync-get-dataset-items waits for the run and returns items directly, capped at 5 minutes. For larger batches, start an async run and fetch the dataset when it finishes (as in the examples above).

Use with AI assistants (MCP)

This Actor can be called directly as a tool by AI assistants and agents through Apify's MCP server — ask your assistant to transcribe a video and it runs the Actor for you.

Claude Code / Claude Desktop:

$claude mcp add --transport http apify https://mcp.apify.com?actors=paintedwinter/tiktok-youtube-transcript --header "Authorization: Bearer YOUR_APIFY_TOKEN"

Cursor and other MCP clients: add https://mcp.apify.com?actors=paintedwinter/tiktok-youtube-transcript as a streamable-HTTP MCP server with your Apify token as the Authorization bearer header.

Your assistant then gets a tiktok-youtube-transcript tool: pass video URLs, get transcripts back in the conversation.

FAQ

Does it work on videos without captions? Yes. It transcribes the audio directly, so it returns text whenever the video has speech — captions are not required.

How long does a video take? Transcription takes about 5–10 seconds per 10 minutes of video. Metadata-only is a few seconds.

Is there a length limit? Transcription is limited to 10 minutes per video; longer videos fail with a clear error and are not charged. Metadata-only mode has no length limit.

Which languages does it support? All major languages are detected automatically. Each result reports the detected language.

Can it do a whole profile or channel? Yes — put the profile/channel/playlist URL in Profiles / channels / playlists and its newest videos are pulled in automatically (set how many with Videos per source, optionally filtered by date or views).

Can I search by keyword? On YouTube, yes — add your query to YouTube search queries. TikTok doesn't expose keyword search, so for TikTok use profile URLs instead.

Does it support Instagram? Not yet. Instagram URLs currently return success: false — free of charge.

Is it legal? This Actor processes publicly available videos only — the same content anyone can view without logging in. It uses no login credentials and collects no private data. You are responsible for using the output in line with the source platform's terms and the laws that apply to you.

---

Built and maintained by Painted Labs.