Audio & Video Transcriber

Turn any audio or video URL into text. Podcasts, interviews, lectures, webinars, meeting recordings, MP3 and MP4 files, and pages from 1,700+ supported sites.

Returns clean text, readable paragraphs, timestamped segments, and ready-to-use SRT and WebVTT subtitles.

Built for long recordings

Most transcription tools quietly fall over on anything longer than a few minutes, because commercial speech-recognition APIs charge by the minute and multi-hour audio gets expensive fast. This actor is designed the other way round.

A real example, start to finish: a 2-hour podcast episode → 25,043 words across 2,577 timestamped segments, correctly aligned to the full 7,200 seconds.

Quick start

{
  "mediaUrls": [
    "https://traffic.libsyn.com/secure/atpfm/atp500.mp3",
    "https://example.com/interview.mp4"
  ],
  "includeParagraphs": true,
  "includeSrt": true
}

Leave language empty and it detects the language itself — detection is accurate and this is usually the right choice. Set it explicitly only when you know the language and the audio is noisy.

What you get back

Field	Description
transcript	Full transcript as clean, continuous text
paragraphs	The same text split into paragraphs on natural pauses — far easier to read
segments	Timestamped segments (start, end, duration, text)
srt	Ready-to-use SRT subtitle content
vtt	Ready-to-use WebVTT subtitle content, for HTML5 players
language · languageProbability	Detected language and confidence
durationSeconds	Length of the recording
wordCount	Words in the transcript
processingSeconds	How long this item took

What it accepts

• Podcast episodes — direct episode URLs from any host
• Direct media files — MP3, M4A, WAV, OGG, OPUS, FLAC, AAC, MP4, WEBM
• Video and audio pages — 1,700+ sites supported

If a URL plays in a browser, there is a good chance this actor can transcribe it.

What people use it for

• Podcast show notes and SEO — publish transcripts so episodes become searchable text
• Meeting and interview records — searchable text instead of scrubbing through audio
• AI and RAG pipelines — feed spoken content into embeddings, summarisation, or Q&A
• Subtitling — export SRT or WebVTT, then translate
• Accessibility — text alternatives for audio and video
• Research — search across hours of recordings by what was actually said

Pricing

Pay per recording transcribed. You are charged only for transcripts actually delivered — a URL that fails, is private, or contains no speech costs you nothing. Failed items still appear in the dataset with an error field, so nothing disappears silently.

Note that the price is per recording, not per minute. A two-hour podcast costs the same as a two-minute clip.

Notes and limits

• Recordings are capped at 2 hours; longer audio is truncated.
• Paywalled, private, and DRM-protected media cannot be transcribed.
• Accuracy is high for clear speech. Heavy background music, crosstalk, and very poor recordings reduce it — as they do for any speech recogniser.
• Long recordings are processed in windows and stitched back together, so timestamps stay accurate across the whole file.
• This actor reads publicly accessible media only. You are responsible for ensuring you have the right to transcribe the content you submit.

Support

Hit a URL that should work but does not, or need another output format? Open an issue on the Issues tab with the URL you used, and it will usually be fixed quickly.