Speech to Text: Video & Audio Transcription from any URL

Get the transcript of videos and audio as clean text, or as timestamped segments, from a URL:

• Videos and posts: X (Twitter), TikTok, YouTube, LinkedIn, Loom, TED, Dailymotion, Twitch clips, Kick, Streamable, Bilibili and 1,000+ other sites, as yt-dlp reads them.
• Playlists, channels and profiles: their latest videos, as many as you ask for.
• Audio and video files: any direct link to an .mp3, .m4a, .wav, .mp4, .webm and the rest, and audio pages such as SoundCloud.

The speech-to-text runs inside the actor: OpenAI's Whisper model, open source, on Apify's servers. Nothing is sent to an outside AI service. It detects the language spoken in each video, among about 100. Long videos are transcribed in full, ten minutes at a time.

Every result also carries what the site shows about the video, as far as it shows it: title, description, uploader, date, duration, views, likes, comments, tags and thumbnail.

It costs $0.003 per minute of audio on Apify's paid plans ($0.006 on the free plan) with Whisper base, and $0.008 ($0.016) with Whisper small, counted in started minutes: a 10-minute video is $0.03. Videos without speech are free.

What people use it for

• AI and LLM pipelines: summaries, RAG over videos, training data, notes from talks and calls.
• Content research: what creators, competitors and experts say, as text you can search.
• Repurposing: blog posts, newsletters, quotes, show notes and subtitles from videos.
• Monitoring: what's said about a brand or topic in videos across platforms.

How to transcribe a video

1. Click Try for free; Apify's free plan covers a first run.
2. Paste the URLs of videos, audio files, playlists or channels under Video or audio URLs.
3. Click Start, then download the results as JSON, CSV, Excel or HTML, or read them through the API.

Max videos per playlist or channel sets how many each of those gives. Max minutes per video transcribes, downloads and charges only each video's first minutes. Speech-to-text model picks between Whisper base, fast and good on clear speech, and Whisper small, more accurate on accents, crosstalk, names and languages other than English, and about three times slower. Language fixes the language spoken when you know it; otherwise each video's is detected. Turn on Include timestamps for segments, each line with its start and end.

{
  "urls": [
    "https://x.com/NASA/status/2116256206922616862",
    "https://www.linkedin.com/posts/the-mathworks_2_what-is-mathworks-cloud-center-activity-7151241570371948544-4Gu7",
    "https://www.youtube.com/playlist?list=PLzMcBGfZo4-mP7qA9cagf68V06sko5otr",
    "https://example.com/files/interview.mp3"
  ],
  "maxVideosPerUrl": 5,
  "model": "base",
  "includeTimestamps": true
}

Output

{
  "input": "https://www.linkedin.com/posts/the-mathworks_2_what-is-mathworks-cloud-center-activity-7151241570371948544-4Gu7",
  "url": "https://www.linkedin.com/posts/the-mathworks_2_what-is-mathworks-cloud-center-activity-7151241570371948544-4Gu7",
  "platform": "LinkedIn",
  "id": "7151241570371948544",
  "title": "What Is MathWorks Cloud Center? | MathWorks",
  "durationSeconds": 162,
  "likeCount": 127,
  "language": "en",
  "languageProbability": 1,
  "text": "Cloud Center enables you to create, access, and manage public cloud resources. You can easily launch Windows or Linux machine with MATLAB install that you can access remotely. …",
  "transcribedSeconds": 161.9,
  "isComplete": true,
  "billedMinutes": 3,
  "segments": [{"start": 0, "end": 4.96, "text": "Cloud Center enables you to create, access, and manage public cloud resources."}]
}

A video from a playlist or channel also has playlistTitle. A video stopped short by Max minutes per video, the run's spending limit or its timeout comes with what was transcribed, isComplete: false, its transcribedSeconds and a stoppedBecause, and is charged only for those minutes. A video without speech (music only, or silent), or one that can't be read, comes with its details and a noTranscriptReason, and costs nothing.

Use it as an API

Apify's API starts a run and returns its dataset as JSON; its Python and JavaScript clients, and its Make, Zapier and n8n integrations, work the same way. Long videos take minutes, so start a run and read its dataset when it ends.

curl -X POST "https://api.apify.com/v2/acts/pikoulas~video-transcriber/runs?token=YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"urls": ["https://x.com/NASA/status/2116256206922616862"], "maxMinutesPerVideo": 30}'

AI agents can call it as a tool through Apify's MCP server: https://mcp.apify.com?tools=pikoulas/video-transcriber.

How much does it cost to transcribe video to text?

With Whisper base you pay $0.003 per started minute of audio on Apify's paid plans, and $0.006 on the free plan: an hour-long video is $0.18 ($0.36 on the free plan). Whisper small, which takes about three times the computing, is $0.008 per started minute ($0.016 on the free plan). billedMinutes shows each video's. Videos without speech are free, and so is everything else the run fetches, with one exception: some sites (YouTube often, Bilibili) refuse Apify's servers, and the run then reaches them through Apify's residential proxy, whose traffic is charged at $0.0012 per 100 KB. Only the sound is downloaded where the site serves it apart, and only the minutes you asked for, so an hour of YouTube audio is about 35 MB, $0.42. There are no monthly fees.

The run stops at the spending limit you set, mid-video if need be, and charges only what it transcribed.

Good to know

• Whisper writes what it hears: names, jargon and crosstalk can come out wrong, and Whisper small gets more of them right. Speakers aren't told apart, and sung lyrics over music are mostly left out with the music.
• A run transcribes about 20 minutes of audio a minute with Whisper base at the default 8 GB of memory, and about 7 with Whisper small. More memory is faster, at the same price per minute. The default timeout of an hour covers about 20 hours of audio with base; raise it for more.
• Facebook, Instagram, Vimeo and Reddit aren't read yet. For Instagram Reels, TikTok and podcasts there are dedicated actors: Instagram Reel Transcripts, TikTok Transcripts and Podcast Transcripts. For YouTube videos with captions, YouTube Transcripts is cheaper.
• Private, members-only, paid and age-restricted videos can't be read; the result says why.
• Media protected against copying (DRM), as streaming services serve it, is never read: the actor doesn't get past copy protection, and such a video comes back with a noTranscriptReason saying so.
• A video given twice in one run, by URL and in a post that embeds it say, is given once.
• Files up to 1.5 GB.

FAQ

Is it legal to transcribe videos?

It reads only what sites show publicly, without logging in. Videos are their creators' copyrighted work: a transcript for your own research, analysis or accessibility is often allowed, depending on your country's copyright law, but republishing it may need the rights holder's permission, and personal data in it is protected by laws such as the GDPR and the CCPA. Sites' terms may also limit how you use what you collect. If you're unsure, ask a lawyer; Apify's Is web scraping legal? is a good start.

Not affiliated with YouTube, TikTok, X or LinkedIn; their names are trademarks of their owners.

Something missing or not working?

Open an issue on the Issues tab with the input you used, or ask there for a site or field you need.