Audio Transcriber — Podcasts, Calls & Meetings

Audio Transcriber turns recordings into readable text: podcasts, interviews, sales calls, meetings, lectures, voice notes and webinars. Every passage keeps the moment it was spoken, so you can jump from a line of text back to the second it came from. 90+ languages, files of any length, and nothing to set up — no key, no model choice, no converting files first.

What it is for

Editorial work (a transcript to quote from), meeting and call notes, compliance archives, subtitles, podcast show notes, search over a back catalogue of recordings, and AI agents that need the spoken word as text before they can reason about it.

What you give it	What you get back
A podcast or interview link	Transcript with timed passages
A recorded call or meeting	The same, optionally labelled by speaker
A video file	The audio is pulled out; the picture is discarded
A three-hour recording	One transcript, timings continuous across parts
A recording in any language	Text in that language, or translated to English
chunkForRag: true	Chunks with timecodes, ready to embed

Formats: MP3, M4A, WAV, FLAC, OGG, MP4, MOV, WEBM — anything with sound, from a public URL or an upload.

What data does it return?

Field	Example
source, fileName	https://example.com/episode-12.mp3 · episode-12.mp3
durationSeconds, durationMinutes	461.05 · 7.68
language, model, translatedToEnglish	English · fast · false
transcript, wordCount, charCount	the full text · 1665 · 9218
segments	[{"start": 0, "duration": 4.56, "text": "…", "speaker": "Speaker 1"}]
speakerCount, speakers	3 · ["Speaker 1", "Speaker 2", "Speaker 3"]
chunks	[{"index": 0, "startTimecode": "00:00:00", "text": "…"}]
subtitles	a complete SRT or WebVTT file
status, errorCode, errorMessage	ok, or why a file failed

How much does it cost?

Event	Price	When it is charged
Minute transcribed	$0.009	Per started minute, using the built-in key
Minute with your own key	$0.004	Per started minute when you supply a Groq key
Add-on: Speaker labels	$0.006	Per started minute, only when speaker labels are on
File processed	$0.002	Per file downloaded and prepared
Video link resolved	$0.09	Per YouTube, TikTok, Instagram, X, Vimeo or other video page whose audio has to be pulled through a residential exit node. Direct file links never pay it

A failed file comes back as an error row and is never charged. Larger monthly plans get 10–30 % off.

Bulk export

This Actor is built for bulk jobs — put hundreds of links into one run, or call it from the API on a schedule. There is no fee per run: you pay per started minute of audio plus $0.002 per file, and $0.09 only for a video page (not a direct file link) whose audio has to be fetched through a residential exit node. Files that fail are returned as error rows and are free.

⬇️ Input

{
  "urls": ["https://example.com/episode-12.mp3"],
  "quality": "accurate",
  "includeSegments": true
}

Speaker labels for a two-person interview:

{ "urls": ["https://example.com/interview.m4a"], "diarize": true, "speakerCount": 2 }

Every input field

Field	What it does
urls	Public links to audio or video files. Several at a time is fine.
file	An uploaded file instead of a link.
quality	fast for speed, accurate for difficult audio and non-English speech.
language	Name the spoken language instead of letting the model guess — faster and safer on short clips.
translateToEnglish	Write the text in English whatever was spoken.
diarize, speakerCount	Split the transcript by speaker. Give the count when you know it.
vocabularyHint	Names, brands and jargon, so they come out spelled your way.
includeSegments	Time-stamped passages alongside the full text.
subtitleFormat	srt or vtt to get a ready subtitle file.
apiKey	Your own Groq key — cuts the per-minute price by more than half.
maxConcurrency	Files processed in parallel.
maxFileSizeMb, timeoutPerFileSecs	Guard rails for very large or very slow downloads.

⬆️ Output

{
  "source": "https://example.com/episode-12.mp3",
  "fileName": "episode-12.mp3",
  "durationMinutes": 7.68,
  "language": "English",
  "transcript": "Welcome back to the show…",
  "wordCount": 1665,
  "segments": [{ "start": 0, "duration": 4.56, "text": "Welcome back to the show" }],
  "status": "ok"
}

Use cases

Podcasts and shows

A transcript to pull quotes from, plus show notes and a searchable archive of every past episode.

Meetings and calls

Minutes that cite the minute something was said, and speaker labels so the actions land on the right person.

Lectures and webinars

Hours of talk turned into text you can search, instead of scrubbing a video for one sentence.

Compliance and records

A written record of calls kept alongside the audio, with timings that point back into the file.

Feeding audio into RAG

chunkForRag returns overlapping chunks with timecodes, sized for embeddings — no post-processing needed.

🤖 For AI agents and LLM apps

Compact reference for agents calling this Actor through the Apify MCP server or the Apify API (lergassy/audio-transcriber).

Purpose: turn a recording into text with time-stamped passages, optionally split by speaker.

Minimal input:

{ "urls": ["https://example.com/audio.mp3"] }

Behaviours an agent should know:

• Billing is per started minute of audio, not per row. A 40-minute file costs the same whether you keep the segments or not.
• durationMinutes on the row tells you what the run actually cost.
• Speaker labels are an extra per-minute charge. Turn diarize on only when who-said-what matters.
• Passing speakerCount when you know it makes the split noticeably cleaner.
• language is detected automatically, but naming it removes the guess on short or noisy clips.
• A file that cannot be downloaded or decoded returns a status: "error" row with the reason, free of charge — check status before reading transcript.
• Video files work: the audio track is extracted and the picture discarded.

Use via API

Run this Actor from your own code or pipeline. Get your token in Apify Console → Settings → API & Integrations. Every input field below has the same name as in the JSON tab of the input form.

Python

from apify_client import ApifyClient

client = ApifyClient("<YOUR_API_TOKEN>")
run = client.actor("lergassy/audio-transcriber").call(run_input={
    "urls": [
        "https://raw.githubusercontent.com/lergassy/apify-actor-assets/main/audio-samples/conversation-4min.ogg"
    ],
    "quality": "fast",
    "translateToEnglish": False,
    "diarize": False,
    "includeSegments": True,
    "chunkForRag": False,
    "chunkSize": 1200,
    "chunkOverlapSeconds": 0,
    "subtitleFormat": "none",
    "maxConcurrency": 3,
    "maxFileSizeMb": 500,
    "timeoutPerFileSecs": 600
})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item)

JavaScript

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<YOUR_API_TOKEN>' });
const run = await client.actor('lergassy/audio-transcriber').call({
    "urls": [
        "https://raw.githubusercontent.com/lergassy/apify-actor-assets/main/audio-samples/conversation-4min.ogg"
    ],
    "quality": "fast",
    "translateToEnglish": false,
    "diarize": false,
    "includeSegments": true,
    "chunkForRag": false,
    "chunkSize": 1200,
    "chunkOverlapSeconds": 0,
    "subtitleFormat": "none",
    "maxConcurrency": 3,
    "maxFileSizeMb": 500,
    "timeoutPerFileSecs": 600
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items);

cURL (runs the Actor and returns the results in one call)

curl -X POST "https://api.apify.com/v2/acts/lergassy~audio-transcriber/run-sync-get-dataset-items?token=<YOUR_API_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{"urls": ["https://raw.githubusercontent.com/lergassy/apify-actor-assets/main/audio-samples/conversation-4min.ogg"], "quality": "fast", "translateToEnglish": false, "diarize": false, "includeSegments": true, "chunkForRag": false, "chunkSize": 1200, "chunkOverlapSeconds": 0, "subtitleFormat": "none", "maxConcurrency": 3, "maxFileSizeMb": 500, "timeoutPerFileSecs": 600}'

Limitations, stated plainly

• Audio quality decides accuracy. Overlapping speech, heavy accents and phone-line compression cost accuracy no model recovers fully.
• Speaker labels are labels, not identities. You get "Speaker 1" and "Speaker 2", never names — mapping them to people is your step.
• Timings are passage-level, not word-level.
• Very large files need time. Raise timeoutPerFileSecs for multi-hour recordings, and expect a long run.
• Links must be public. A file behind a login or an expiring signed URL cannot be fetched.

Integrations

Runs from the Apify API and the Python and JavaScript clients, and from n8n, Make or Zapier through the Apify app. Point a webhook at the dataset to push transcripts into Notion, Google Docs or your own database, or let an agent call it through the Apify MCP server.

❓ FAQ

Do I need an API key or an account anywhere?

No. Recognition runs with a built-in key. Supply your own Groq key only if you want the cheaper per-minute rate.

Which languages are supported?

Around 90, including Russian, Indonesian, Spanish, German, French, Japanese and Chinese. Use accurate for non-English audio.

Can it handle a three-hour file?

Yes. Long files are processed in parts and the timings stay continuous across them.

Does it work with video?

Yes. Give it an MP4, MOV or WEBM link and the audio track is pulled out for you.

What happens if a file is unreachable?

You get an error row with the reason and no charge for that file. The other files in the run are still transcribed.

How accurate is it?

It runs Whisper large v3. On clean speech it is close to a careful human first pass; on overlapping or noisy audio it is not.

Can I get speaker names?

No — you get "Speaker 1", "Speaker 2" and so on. Naming them is your step.

How do I make it spell product names correctly?

Put them in vocabularyHint. The hint is given to the model before it starts.

You might also like

Actor	What it does
Speech to Text	The same engine, framed as a speech-to-text API
SRT Subtitle Generator	Ready subtitle files from audio or video
YouTube Transcript Scraper	Captions straight from YouTube, no transcription needed
Document Text Extractor	The same idea for PDFs and Word files

Also known as

People look for this Actor as an audio transcriber, audio to text converter, podcast transcription API, meeting transcription tool, interview transcription service and a speech recognition API.

Notes

Recognition runs on Whisper large v3 with a built-in key, so there is nothing to sign up for. Choose fast (Whisper large v3 turbo, several times quicker) for everyday work, or accurate for accents, noise and translation; both cost the same per minute. A vocabulary hint helps with names and jargon.