Audio & Video Transcription with Speaker Diarization — by Brainiall

Turn public audio and video URLs into clean transcripts with timestamps, optional speaker identification, and ready-to-save SRT or WebVTT captions.

Add one URL or a batch. Each file becomes a structured dataset item that can flow directly into content workflows, search indexes, meeting archives, research pipelines, and accessibility tooling.

What it does

• Audio and video URL transcription — process one file or a whole batch in a single run.
• Speaker diarization — identify who spoke when and label each timestamped transcript segment.
• Structured timestamps — receive full text plus machine-readable segments with start and end times.
• Captions on demand — optionally include SRT, WebVTT, or both with no second conversion step.
• Resilient batch processing — temporary rate limits and service errors are retried; one invalid file becomes an error item without stopping the rest of the run.

Powered by the Brainiall speech service (api.brainiall.com) — built for dependable batch workflows.

Who it's for

• Media teams — create editable transcripts and captions for interviews, podcasts, webinars, and recorded video.
• Researchers and analysts — preserve who said what and when across recorded conversations.
• AI builders — feed timestamped, speaker-aware text into search, summarization, RAG, and agent pipelines.
• Support and sales teams — turn call recordings into structured data ready for downstream analysis.
• Accessibility workflows — generate SRT and WebVTT caption content from hosted media.

Input

{
    "mediaUrls": [
        { "url": "https://example.com/interview.mp3" },
        { "url": "https://example.com/product-demo.mp4" }
    ],
    "diarization": true,
    "includeSrt": true,
    "includeVtt": false
}

Field	Description
mediaUrls	Public HTTP(S) URLs of audio or video files. Each URL produces one dataset item.
diarization	Identify speakers and add labels to transcript segments. Defaults to true.
includeSrt	Include a ready-to-save SRT caption string. Defaults to false.
includeVtt	Include a ready-to-save WebVTT caption string. Defaults to false.

The Actor downloads each hosted media file into memory before processing it. Files must be no larger than 100 MB; larger downloads produce an error item without stopping the batch.

Output

One dataset item per media URL:

{
    "media_url": "https://example.com/interview.mp3",
    "status": "succeeded",
    "text": "Welcome to the show. Thanks for having me.",
    "language": "en",
    "duration_seconds": 6.8,
    "billable_minutes": 1,
    "segments": [
        {
            "start": 0.0,
            "end": 2.7,
            "text": "Welcome to the show.",
            "speaker": "SPEAKER_00"
        },
        {
            "start": 3.1,
            "end": 6.8,
            "text": "Thanks for having me.",
            "speaker": "SPEAKER_01"
        }
    ],
    "speaker_turns": [
        { "start": 0.0, "end": 2.9, "speaker": "SPEAKER_00" },
        { "start": 3.0, "end": 6.8, "speaker": "SPEAKER_01" }
    ],
    "srt": "1\n00:00:00,000 --> 00:00:02,700\n[SPEAKER_00] Welcome to the show.\n\n2\n00:00:03,100 --> 00:00:06,800\n[SPEAKER_01] Thanks for having me."
}

If one URL is invalid or cannot be processed, it still produces a useful, uncharged error item while the remaining URLs continue:

{
    "media_url": "https://example.com/missing.mp3",
    "status": "failed",
    "error": {
        "type": "media_rejected",
        "message": "The media could not be processed (HTTP 400)."
    }
}

Export results as JSON, CSV, or Excel, or consume them through the Apify API.

Pricing

You pay per started minute successfully transcribed through the minute-transcribed event. For example, a 75-second successful file uses two transcription-minute events. Failed inputs and incomplete results are not charged. There is no batch minimum; the current per-event price is shown on the Actor run page before you start.

Why this Actor

Most transcription steps leave you with a wall of text. This Actor returns an automation-ready record: the complete transcript, timestamped segments, speaker attribution, raw speaker turns, and optional caption formats together. Batch isolation and short retries keep one troublesome URL from wasting a full run.

Tips

• Use direct, publicly reachable media-file URLs rather than links to pages containing an embedded player.
• Turn off speaker identification for single-speaker recordings when you only need text and timestamps.
• Request only the caption formats you need to keep dataset items compact.
• Keep each hosted file at or below 100 MB.

---

Built and maintained by Brainiall — production AI APIs for speech, documents, vision, and the web.