AI Audio to Text Transcriber

Turn audio files into clean, verbatim text. Give it a list of public audio URLs and it returns a full transcript for each one — no API key, no setup, no per-vendor billing to manage. You pay per minute of audio transcribed.

What it does

• Accepts a list of public audio file URLs (MP3, MP4, M4A, WAV, WEBM, OGG, FLAC)
• Downloads each file to temporary storage (max 25 MB per file)
• Returns the full verbatim transcript plus the measured audio duration
• Processes up to 3 files concurrently for faster batch runs
• Saves one dataset record per file, including error records for files that fail

Use cases

• Podcast indexing and search
• Meeting recording notes
• Compliance and call-center transcription
• Generating training data for NLP models
• Subtitles and captions for video content
• Multilingual content analysis

Input

Field	Type	Required	Description
audioUrls	Array	Yes	Public audio file URLs to transcribe
language	String	No	ISO 639-1 hint (e.g. en, es, ja). Omit for auto-detect.
maxItems	Integer	No	Maximum files to transcribe per run. Default: 15.

Supported audio formats: MP3, MP4, M4A, WAV, WEBM, OGG, FLAC Max file size: 25 MB per file

Example input

{
  "audioUrls": [
    "https://example.com/podcast-episode-1.mp3",
    "https://example.com/meeting-recording.wav"
  ],
  "language": "en",
  "maxItems": 10
}

Output

One dataset record per audio file.

Field	Type	Description
sourceUrl	String	Original audio file URL
transcript	String	Full verbatim transcription text
language	String	The language hint you supplied, or null if auto-detected
durationSeconds	Number	Measured audio duration in seconds
billedMinutes	Number	Whole audio minutes charged for this file
model	String	Transcription model used
transcribedAt	String	ISO timestamp
status	String	success or error
errorMsg	String	Error description on failure, null on success

Example output record

{
  "sourceUrl": "https://example.com/podcast-ep1.mp3",
  "transcript": "Welcome to today's episode. Today we're discussing the future of AI...",
  "language": null,
  "durationSeconds": 1823.4,
  "billedMinutes": 31,
  "model": "gpt-4o-mini-transcribe",
  "transcribedAt": "2026-08-03T12:00:00Z",
  "status": "success",
  "errorMsg": null
}

Requirements

• Public audio URLs — Files must be publicly accessible without authentication.

That's it. No API key to obtain, and no separate vendor bill.

Pricing

$0.10 per run start + $0.015 per minute of audio transcribed.

Minutes are rounded up per file, with a one-minute minimum — a 20-second clip bills one minute, a 10-minute podcast bills ten. A file that fails to download or transcribe bills the same one-minute minimum.

There is nothing else to pay: transcription costs are included in the per-minute price.

Error handling

Files that fail to download or transcribe are not dropped — the actor saves an error record to the dataset with status: "error" and a descriptive errorMsg, so your dataset always has one row per input URL for easy reconciliation.

Common errors:

• HTTP 429 — rate limit exceeded (retry with fewer files)
• File exceeds 25 MB limit — source file too large
• Download timed out — URL not reachable within 60 seconds
• ffprobe failed to read audio duration — file is not decodable audio, or is corrupt