Audio and Video File Transcriber (timestamps, 90+ languages)

Turn audio and video files into accurate text with timestamps. Paste direct file links or upload a file. The language is detected for you, and you can translate the speech into English. It runs on OpenAI's open-source Whisper model, so you need no API key and no account anywhere else.

Price: US$0.006 per audio minute, plus a US$0.001 run fee. A 30-minute recording costs US$0.181.

What people use it for

• Meetings and interviews: searchable text from recordings, with timestamps to jump back to.
• Podcasts and webinars: transcripts for show notes, blog posts and SEO.
• Research: transcribe focus groups and field recordings in many languages.
• AI pipelines: feed transcripts to an LLM for summaries, chapters or Q&A.
• Accessibility: text versions of audio content.

Need subtitle files instead? Use the Subtitle Generator below. It makes SRT and VTT directly.

Example output

Real output from the sample file (a 117-second public-domain recording):

fileName	durationSec	billedMinutes	language	text
Wikipedia_-Chojnów(spoken_by_AI_voice).mp3	117.0	2	en	Chajno is small city in Poland. As of the year 2000, 14,900 people live in the city...

{
  "sourceUrl": "https://upload.wikimedia.org/wikipedia/commons/2/2a/Wikipedia_-_Chojn%C3%B3w_%28spoken_by_AI_voice%29.mp3",
  "fileName": "Wikipedia_-_Chojnów_(spoken_by_AI_voice).mp3",
  "durationSec": 117.0,
  "billedMinutes": 2,
  "language": "en",
  "languageProbability": 0.998,
  "model": "base",
  "task": "transcribe",
  "text": "Chajno is small city in Poland. As of the year 2000, 14,900 people live in the city...",
  "segments": [
    {"start": 0.0, "end": 4.16, "text": "Chajno is small city in Poland. As of the year 2000,"},
    {"start": 4.8, "end": 12.08, "text": "14,900 people live in the city. The city is on the river Scora."}
  ],
  "transcribedAt": "2026-10-27T09:00:00+00:00"
}

The real place name is Chojnów. Like every speech model, Whisper spells unusual names by sound. The "accurate" setting gets more of them right.

Input

Field	What it does	Default
mediaUrls	Direct links to audio or video files, up to 100. MP3, WAV, M4A, OGG, FLAC, MP4, MOV, WEBM, MKV and more. 1 GB per file.	1 sample file
uploadedFile	Or upload one file from your computer.	none
model	tiny (fast), base (balanced) or small (accurate).	base
language	auto, or a code such as en, es, de, hi, ja. Setting it improves accuracy.	auto
task	transcribe keeps the spoken language. translate outputs English.	transcribe
wordTimestamps	Add a start and end time for every word.	false
skipSilence	Skip silence and music-only parts. Faster, and fewer made-up words.	true
maxMinutesPerFile	Longer files are skipped without charge. A safety limit on cost.	180

Pricing

Event	Price
Run fee, once per run (per 2 GB of memory; the default is 2 GB)	US$0.001
Audio minute, fast or balanced model	US$0.006
Audio minute, accurate model	US$0.012

Minutes are counted per file and rounded up. Examples: a 1-minute clip costs US$0.007. A 30-minute file costs US$0.181. A 60-minute file on the accurate model costs US$0.721. Files that fail to download or decode are listed with the reason and not charged.

Limits

• Direct file links only. Links to YouTube, TikTok, Instagram, Facebook, X, Vimeo, Spotify, SoundCloud and similar pages are refused. Only process files you have the right to use.
• 100 files per run, 1 GB and 600 minutes per file.
• Speed at the default 2 GB memory: about 1 to 3 minutes of processing per 10 audio minutes on the balanced model. Raise the memory for faster runs. The run fee scales with memory.
• No speaker labels (who said what).

FAQ

Which languages work? Whisper supports about 99 languages. Accuracy is best for English, Spanish, German, French, Portuguese, Italian, Japanese and other widely spoken languages.

Is my file stored? It is downloaded to the run's temporary disk, processed and deleted. The transcript is saved in your own Apify dataset.

Why do transcripts sometimes contain words nobody said? Speech models can invent text during long silences or music. Keep skipSilence on to reduce this.

Can an AI agent call it? Yes. Pass file links, get compact JSON back, and pay per minute.

More tools from the same developer

Same style: clean JSON output, pay only for results, no charge on failed items.

Tool	What it does
Bulk Keyword Search Volume	Google Ads search volume, CPC, competition and 12-month trend for up to 10,000 keywords
Google Trends API	Interest over time, by region, and rising queries and topics for up to 500 keywords
Google SERP Results	Organic results, featured snippets and People Also Ask for any keyword, country and device
Google AI Overview Tracker	Which keywords show an AI Overview, what it says, and whether it cites your site
Subtitle Generator	Ready-to-use SRT and WebVTT subtitles from audio and video files
Podcast Transcriber	Timestamped transcripts of podcast episodes from any RSS feed or Apple Podcasts link
Image OCR	Text from images in 20 languages, with confidence scores
PDF to Text and Markdown	Clean text and Markdown from PDFs, with OCR for scanned pages
Bulk PageSpeed Insights Audit	Core Web Vitals and Lighthouse scores for many pages, mobile and desktop