Video & Audio Transcriber

Give it a public video or audio URL and it returns accurate text with segment and word-level timestamps, plus ready-to-use SRT, VTT, and TXT files. It detects the spoken language automatically. Built for people who need captions, searchable transcripts, or source text to repurpose into clips, articles, or show notes.

How it works

The actor downloads your media, extracts the audio track with ffmpeg, and sends it to OpenAI's Whisper on your own API key. The timestamps and subtitle files come straight from the model's segment and word data, so timing lines up with the actual speech.

Input

Field	Required	Notes
mediaUrl	one of these	Public URL to a single video or audio file (mp4, mov, mp3, wav, m4a, webm, and similar).
mediaUrls	one of these	An array of public URLs to transcribe in one run — one dataset row per URL.
language	no	ISO code of the spoken language, or auto to detect it. Defaults to auto.
wordTimestamps	no	Return per-word start/end times. Useful for karaoke-style captions. On by default.
outputFormats	no	Which files to generate: any of srt, vtt, txt. Defaults to srt and vtt.
openaiApiKey	yes	Your OpenAI (Whisper) key. Kept private and used only for this run.

There are two advanced fields if you need them: model (defaults to whisper-1) and baseUrl, which points the actor at any OpenAI-compatible endpoint instead of api.openai.com.

Output

One dataset record per run. It includes the detected language, the full text, segments with start/end times, and words when word timestamps are enabled, along with wordCount, segmentCount, and durationSeconds. Each requested subtitle file is saved to the key-value store and referenced by srtKey/srtUrl, vttKey/vttUrl, and txtKey/txtUrl.

Example

{
  "mediaUrl": "https://example.com/podcast.mp3",
  "language": "auto",
  "wordTimestamps": true,
  "outputFormats": ["srt", "vtt", "txt"],
  "openaiApiKey": "sk-..."
}

Pricing

$0.02 per minute of audio transcribed, plus $0.0000125 each time a run starts. Flat rate — no volume tiers, no plan gates — and you are charged only for minutes actually transcribed. A 40-minute podcast costs $0.80; a 3-minute clip costs $0.06.

That is a flat rate on every Apify plan, which is unusual in this market: transcription is normally billed per second on a tier that only improves once you are on a high-volume plan. Here it is $0.02 per minute from your very first run and it does not change with your plan. Word-level timestamps, SRT/VTT/TXT output and language detection are part of the same per-minute charge rather than per-second add-ons.

You bring your own OpenAI key, so Whisper usage is billed to you by OpenAI separately and is not marked up here.

Notes

The mediaUrl has to be directly downloadable. Pages that require login or stream behind a player won't work, so point it at the raw file. Long files take longer and cost more since billing is per minute of audio. Batch runs via mediaUrls are billed the same way — the total is the sum of the minutes across every file, not a per-file fee.