Audio & Video Transcriber: Transcription, Speech to Text, SRT avatar

Audio & Video Transcriber: Transcription, Speech to Text, SRT

Pricing

from $5.00 / 1,000 transcribed audio minutes

Go to Apify Store
Audio & Video Transcriber: Transcription, Speech to Text, SRT

Audio & Video Transcriber: Transcription, Speech to Text, SRT

Audio transcription with Whisper large-v3, $6 per 1,000 minutes: transcribe audio and video (MP3, WAV, M4A, MP4, MOV, WebM) to text, timestamped segments and SRT/VTT subtitles. Speech to text in 99 languages incl. Ukrainian; podcasts, meetings, interviews; long files split. No API key.

Pricing

from $5.00 / 1,000 transcribed audio minutes

Rating

0.0

(0)

Developer

Andriy Kuzmenko

Andriy Kuzmenko

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

5 hours ago

Last modified

Features

Limited permissions

Share

Audio & Video Transcriber: Speech to Text, SRT/VTT Subtitles

Turn audio and video files into text, timestamped segments and ready-to-use SRT / VTT subtitles with OpenAI's Whisper large-v3 model - 99 languages, automatic language detection, files of any length.

  • Any audio or video format. MP3, WAV, M4A, AAC, OGG, OPUS, FLAC, MP4, MOV, MKV, WebM, AVI… The speech track is extracted automatically.
  • Long recordings. Meetings, lectures, podcasts and interviews are split into 10-minute parts and stitched back with continuous timestamps.
  • Strong non-English support. Whisper large-v3 handles Ukrainian, Polish, German, Spanish, Russian and 90+ more languages; a vocabulary hint fixes the spelling of names and terms.
  • Subtitles included. SRT and WebVTT files with download links, plus segments as JSON.

Why this Actor

This ActorTypical alternatives in the Apify Store (Oct 2026)
Price$6 per 1,000 minutes (36 cents an hour; $5 on Gold+), timestamps and SRT/VTT included$6-28 per 1,000 minutes
ModelWhisper large-v3 - the most accurate Whisper model - for every fileOften smaller, cheaper models, or large-v3 only in a premium mode
SubtitlesSRT + VTT + timestamped JSON segments in every runOften text only, or subtitles as a paid extra
No API keyNothing to sign up for; pay per minute through ApifySome ask for your own OpenAI / Groq key
Long filesHours-long recordings are split and stitched with continuous timestampsTools that send the file to a Whisper API in one request hit its 25 MB limit

Input

{
"sources": [{ "url": "https://example.com/interview.mp3" }, { "url": "https://example.com/webinar.mp4" }],
"language": "uk",
"quality": "accurate",
"vocabulary": "Lviv, Kyivstar, PrivatBank",
"subtitles": true
}
FieldMeaning
sourcesURLs of audio or video files. Files in your own Apify key-value store work too: set the store's access to "Anyone with ID can read" and use https://api.apify.com/v2/key-value-stores/<store ID>/records/<key> (the Actor runs with limited permissions and does not use your token).
languageISO code (en, uk, de…) or empty for automatic detection
qualityaccurate (Whisper large-v3, default) or fast (large-v3-turbo) - same price
crossCheckRecognise the audio again with the other model and list where the two disagree (see below). Each minute is charged twice
vocabularyNames and terms to spell correctly; Ukrainian city names (Kyiv, Lviv, Kharkiv…) are built in when language is en, uk or ru
subtitlesSave SRT and VTT files (free)

Output

{
"url": "https://example.com/interview.mp3",
"fileName": "interview.mp3",
"language": "Ukrainian",
"durationSeconds": 1834.2,
"billedMinutes": 31,
"model": "whisper-large-v3",
"text": "Добрий день! Сьогодні ми поговоримо про…",
"segments": [{ "start": 0.0, "end": 4.6, "text": "Добрий день!" }],
"srtUrl": "https://api.apify.com/v2/key-value-stores/.../records/0001-interview.srt?signature=...",
"vttUrl": "https://api.apify.com/v2/key-value-stores/.../records/0001-interview.vtt?signature=..."
}

A file that cannot be downloaded or has no audio track gets an error field instead and is not charged.

Missed speech (gaps)

Whisper sometimes skips a sentence or two: the transcript reads smoothly, but a few seconds of speech are not in it, or are a single word («The» for seven seconds). Every transcript is checked for this, free of charge. Each stretch of 5 seconds or more with no text, or almost none, is measured with ffmpeg; where it holds sound and not silence, it is recognised again on its own. The item then lists these places:

"gaps": [
{ "start": 18.8, "end": 25.4, "status": "filled", "text": "next shipment of cement and steel rods is planned..." },
{ "start": 61.0, "end": 70.2, "status": "unclear", "text": "" }
]

filled: the speech was found and is now in text, segments and the subtitles. unclear: there is sound but no clear speech was found (music, noise, crosstalk), worth listening to. No gaps field: nothing was missed.

Cross-check: where to listen again

Wind, music, a bad phone line or several people talking at once break numbers and short words first, and the two Whisper models break them differently. With "crossCheck": true the same audio is recognised by the second model as well, and the item gets the places where the two transcripts disagree. A real example, a weather report recorded outdoors in a storm:

"crossCheck": {
"model": "whisper-large-v3-turbo",
"agreement": 0.884,
"differences": [
{ "start": 0.0, "end": 7.0, "text": "gusts of wind of 17 even 18", "other": "just a room of 70 even 80", "numbers": true },
{ "start": 21.0, "end": 28.0, "text": "with drought in", "other": "throughout", "numbers": false }
],
"text": "…the full transcript of the second model…"
}

The forecast for that day was gusts of 60-70 mph, up to 80, so the second model got the figures right, and «throughout the morning» fits the sentence; in turn it misheard «gusts of wind of». numbers: true marks the differences with figures in them, the ones that matter most in minutes, interviews and calls. Times come from the main transcript. Where both models agree they can still both be wrong, so this narrows the listening down, it does not replace it. billedMinutes shows the doubled count.

Pricing

EventPriceWhat counts
audio-minute$6 per 1,000 minutes ($7 on the free plan, $5 on Gold and higher)every started minute of transcribed audio (at least 1 per file); twice with crossCheck

Plus Apify's standard start fee of $0.05 per 1,000 runs. Subtitles are free. The run transcribes only as many minutes as fit under your maximum charge per run and marks the item with a warning.

Privacy

The speech track is sent to Groq's speech-to-text API, which runs the Whisper model; see Groq's privacy policy. This Actor keeps files only in your own Apify run storage.

Guides

Open source: github.com/AndriyVK/-apify-document-tools

Use with AI agents and code

  • MCP (Claude, Cursor, VS Code, any MCP client): add https://mcp.apify.com?tools=ai_admin/audio-transcriber as an MCP server - the agent gets one tool with this Actor's input schema and pays per use through your Apify account.
  • One HTTP call that returns the results:
curl -X POST "https://api.apify.com/v2/acts/ai_admin~audio-transcriber/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
-H "Content-Type: application/json" \
-d '{"sources": [{"url": "https://api.apify.com/v2/key-value-stores/FloomuI2DrQCJazY8/records/speech-en.mp3"}], "language": "en"}'
  • Agentic payments: the Actor is enabled for Apify's agentic payments (x402), so autonomous agents can pay per run.

Limits

  • YouTube and other streaming pages are not supported - provide a direct link to the media file.
  • Maximum download size is 200 MB per file.
  • Speaker labels (diarization) are not included.