Audio Transcription API: Subtitles, Speakers & RAG

Batch transcribe direct audio and video URLs or uploaded files into timestamped text, SRT subtitles, and VTT subtitles. Add speaker labels, custom glossary terms, optional sensitive-data redaction, or citation-ready RAG chunks. No transcription API key is required.

Supported inputs

• Direct public or signed HTTP(S) links to audio/video files.
• Files uploaded through Apify.
• Up to 100 files per run, including MP3, M4A, WAV, FLAC, OGG, Opus, WebM, MP4, MOV, and MKV.

This Actor processes direct media files. It does not extract media from YouTube, TikTok, Instagram, Google Drive, Dropbox, or other social and webpage URLs.

Quick start

Paste a direct media URL or upload a file. This public sample is ready to run:

{
  "mediaUrls": [
    "https://static.deepgram.com/examples/interview_speech-analytics.wav"
  ],
  "mode": "conversation",
  "language": "auto",
  "glossary": ["speech analytics", "KPIs"],
  "outputFormats": ["txt", "srt", "vtt", "json"],
  "createRagChunks": true
}

Use dryRun: true to validate files and estimate rounded-up billable minutes without transcribing or charging transcription events.

What you get

Each successful input produces:

• Timestamped transcript text with confidence and detected language.
• Speaker labels in conversation mode.
• Downloadable TXT, Markdown, JSON, SRT, and VTT artifacts.
• Low-confidence ranges and glossary coverage to help review output.
• Optional RAG chunks with source URL, timestamp range, speaker IDs, token estimate, and content hash.

Example transcript:

[Speaker 0] Another big problem in the speech analytics space is that customers are often overwhelmed by the number of KPIs they can track.

Example downloadable SRT subtitle content:

1
00:00:00,000 --> 00:00:04,800
Speaker 0: Another big problem in the speech analytics space

2
00:00:04,800 --> 00:00:08,600
is that customers are often overwhelmed by the number of KPIs.

Artifact URLs are returned on every successful Dataset result row, including srtUrl and vttUrl for direct download.

Modes and options

Option	Use it for
standard	Single-speaker recordings or basic transcription.
conversation	Interviews, meetings, podcasts, and calls where speaker labels matter.
glossary	Product names, acronyms, people, brands, and specialist terms.
redact	PII, PCI, PHI, or number redaction.
createRagChunks	Citation-ready transcript chunks for retrieval workflows.
multichannel	Calls recorded with participants on separate audio channels.

language accepts auto, multi for code-switching, or a supported language code. PII, PCI, and PHI entity redaction is currently English-only; number redaction supports additional languages.

Pricing

Files are rounded up to the next full minute individually. You pay only after a requested transcription completes and its requested outputs are written.

Processing	Current price
Standard transcription	$0.015 per rounded-up minute
Conversation transcription with speaker labels	$0.022 per rounded-up minute
Sensitive-data redaction add-on	+$0.004 per rounded-up minute
Actor start	$0.00005 per run

For example, a 96.6-second standard recording is billed as 2 minutes: $0.030, plus the $0.00005 Actor start charge.

Results and exports

The default Dataset contains one result per input. A successful row includes the transcript, timestamped segments, speaker count, confidence, billing details, artifact URLs, and the RAG Dataset link when enabled. Failed inputs return a structured error without transcription-minute charges, while other inputs in the batch continue.

Use Apify's Dataset export to download result rows as JSON, CSV, Excel, or other supported formats.

Privacy and data handling

Files are downloaded into an isolated temporary directory and deleted after processing. Signed URL query strings are used for download but are removed from logs, result rows, Markdown, and RAG provenance.

When privacyMode is enabled, the Actor requests that the transcription provider exclude media from model-improvement programs where supported. Redaction is performed before the transcript is returned, but media is sent to a third-party transcription provider for processing. Apify storage retention and access are controlled by the Actor run owner.

This is not a compliance certification. Customers handling regulated data should evaluate their own privacy, security, and compliance requirements before use.