Audio & Video to Text - Whisper Transcriber with SRT & VTT Subtitles

Transcribe audio and video files to text with OpenAI Whisper: a clean transcript, timestamped segments, ready-to-use SRT and WebVTT subtitle files, optional word-level timestamps, in 99 languages with automatic language detection. Paste links to MP3, WAV, M4A, MP4, MOV, WEBM or MKV files and get text back in minutes. No API key, no OpenAI account, no GPU. You pay $0.008 per audio minute, only for files that were actually transcribed.

What does Audio & Video to Text do?

It downloads each audio or video file you link to, pulls out the sound track, skips silence with voice-activity detection and runs Whisper speech recognition on the speech. For every file you get:

• the full transcript as plain text, split into paragraphs at pauses,
• timestamped segments (start, end, text) sized like real subtitles (max 2 lines of 42 characters),
• SRT, VTT and TXT files saved with download links (plus an optional full JSON file),
• the detected language and its confidence, duration, codec and speech length,
• optional word timestamps (start and end of every word).

Use cases

• AI and RAG pipelines - turn podcasts, webinars, calls and course videos into text an LLM can search, summarize or answer questions about.
• Subtitles and captions - get SRT/VTT files for your own videos, lessons and social clips.
• Podcasts and interviews - show notes, quotes and searchable archives; paste the RSS feed and a schedule transcribes each new episode.
• Meetings and voice notes - text you can skim, search and share.
• Content repurposing - blog posts, newsletters and threads from your recordings.
• Video analysis - pair the transcript with keyframes from our Video Frame Extractor so a vision model sees and hears the video.

How to transcribe a file

1. Click Try for free.
2. Paste direct links to your audio or video files into Audio or video file URLs, one per line.
3. Leave Spoken language on auto, or set a code like en, es, de, ja.
4. Click Start. Open the transcript in the results table, or download the SRT/VTT/TXT files from the links.

Input

Field	What it does	Default
fileUrls	Direct links to audio/video files. Dropbox and Google Drive share links are converted automatically	1 sample (Gettysburg Address, public domain)
language	auto or an ISO 639-1 code (en, es, fr, de, ja, zh, hi...)	auto
outputFormats	Files to save: txt, srt, vtt, json	txt, srt, vtt
wordTimestamps	Add start/end of every word	false
subtitleMaxChars	Longest subtitle cue (2 lines)	84
maxDurationMinutes	Skip longer files (free)	180 (max 300)
maxFileSizeMb	Skip larger files (free)	1000 (max 4000)
downloadTimeoutSecs	Give up on slow downloads (free)	600

{
    "fileUrls": [
        "https://your-bucket.s3.amazonaws.com/podcasts/episode-42.mp3",
        "https://www.dropbox.com/scl/fi/abc123/interview.mp4?dl=0"
    ],
    "language": "auto",
    "outputFormats": ["srt", "vtt", "txt"]
}

Podcast RSS + scheduled new episodes

Give it a podcast's public RSS or Atom feed and it transcribes the episodes' audio files, newest first. Each result row carries the episode metadata next to the transcript: feedUrl, podcastTitle, episodeTitle, episodePublishedAt, episodeGuid and episodeUrl.

Field	What it does	Default
rssFeedUrls	Public podcast RSS/Atom feed URLs	-
maxEpisodesPerFeed	How many of the newest episodes of each feed to look at per run	5
episodeDateFrom	Leave out older episodes: 2026-10-01 or 7 days	-
onlyNewEpisodes	Transcribe (and charge) only episodes not transcribed by an earlier run	false
stateStoreName	Named key-value store in your account that remembers done episodes	kanto-transcriber-state

Scheduled "new episodes only" mode. Turn on onlyNewEpisodes and put the input on an Apify schedule (daily or weekly). Every run reads the feeds, skips the episodes it already transcribed (they produce no row and no charge, just a log line) and transcribes only the new ones. You pay once per episode. Use a different stateStoreName per schedule to keep them apart; delete that store to start over.

{
    "rssFeedUrls": ["https://feeds.npr.org/510318/podcast.xml"],
    "maxEpisodesPerFeed": 3,
    "onlyNewEpisodes": true,
    "stateStoreName": "my-podcast-monitor",
    "outputFormats": ["txt", "srt"]
}

n8n: a Schedule Trigger (every morning) -> the Apify node, operation Run actor and get dataset, actor kantolabs/audio-video-to-text, with the JSON input above -> each new episode arrives as one item with episodeTitle, text and srtUrl; send it to an LLM for show notes, to Notion/Sheets for an archive, or to Slack. A morning with no new episode returns 0 items, so nothing downstream fires. Make: Schedule -> Apify Run an Actor (wait) -> Get dataset items.

Notes: reading feeds is free; you pay only the transcribed minutes. Feeds that fail (404, a web page instead of a feed, a feed without audio files) return a free failed row that says what to fix. Apple Podcasts and Spotify pages are not feeds - use the show's own RSS URL, which every public podcast has. If you leave the sample link in fileUrls while adding a feed, the sample is ignored.

Supported formats: MP3, WAV, M4A, AAC, OGG, OPUS, FLAC, WMA, AMR, MP4, MOV, M4V, WEBM, MKV, AVI, MPEG-TS and anything else ffmpeg can decode. Video files are fine - only the audio track is used.

Output

One dataset item per file. Real output from a cloud test run (text shortened, segments trimmed to 3 of 42):

{
    "fileUrl": "https://archive.org/download/gettysburg_johng_librivox/gettysburg_address_64kb.mp3",
    "fileName": "gettysburg_address_64kb.mp3",
    "fileSizeBytes": 1350656,
    "durationSeconds": 157.884,
    "container": "mp3",
    "audioCodec": "mp3",
    "hasVideo": false,
    "speechSeconds": 144.3,
    "status": "ok",
    "language": "en",
    "languageProbability": 0.999,
    "task": "transcribe",
    "model": "Whisper base",
    "transcribedSeconds": 157.884,
    "truncated": false,
    "billedMinutes": 3,
    "wordCount": 308,
    "segmentCount": 42,
    "text": "This is a Libravox recording. All Libravox recordings are in the public domain. For more information or to volunteer, please visit Libravox.org.\n\nThe Gettysbur...",
    "segments": [
        {
            "id": 1,
            "start": 0.05,
            "end": 1.91,
            "text": "This is a Libravox recording."
        },
        {
            "id": 2,
            "start": 2.39,
            "end": 4.87,
            "text": "All Libravox recordings are in the public domain."
        },
        {
            "id": 3,
            "start": 5.49,
            "end": 9.99,
            "text": "For more information or to volunteer, please visit Libravox.org."
        }
    ],
    "txtUrl": "https://api.apify.com/v2/key-value-stores/dixfgDWP87NhufUSb/records/49d29889-gettysburg_address_64kb.txt?signature=...",
    "srtUrl": "https://api.apify.com/v2/key-value-stores/dixfgDWP87NhufUSb/records/49d29889-gettysburg_address_64kb.srt?signature=...",
    "vttUrl": "https://api.apify.com/v2/key-value-stores/dixfgDWP87NhufUSb/records/49d29889-gettysburg_address_64kb.vtt?signature=...",
    "processingTimeSeconds": 15.1,
    "realTimeFactor": 0.082,
    "error": null
}

The SRT/VTT/TXT links open without an Apify login, so you can pass them straight to a video player, an editor, an LLM or a spreadsheet. A file that could not be transcribed gets a row with "status": "failed" and an error that says how to fix the input - and costs nothing.

How much does it cost to transcribe audio?

Pay per event, no start fee per file, no monthly subscription:

Event	Price	When
Audio minute transcribed	$0.008 per started minute ($8 per 1,000 minutes)	the file was transcribed

Minutes are counted on the file's audio length, rounded up per file with a 1-second grace (a 2 min 10 s file = 3 minutes, a 60.4 s file = 1 minute). Free: refused links, failed downloads, unreadable files, files with no audio track, files with no speech (music or silence only), files over your size/duration limits.

Worked examples:

• A 45-minute podcast episode = 45 minutes = $0.36
• 100 product videos of 1 minute each = $0.80
• A 3-hour webinar = 180 minutes = $1.44

Set Maximum cost per run and the Actor never transcribes more minutes than that budget allows: the last file is cut at the budget (and marked truncated), later files are listed as skipped, free.

Speed and real costs (measured on Apify, October 2026)

Test file	Audio	Run time	Billed	Price
Gettysburg Address (MP3, English)	2 min 38 s	26 s	3 min	$0.024
"Duck and Cover" 1951 film (MP4 video, English)	9 min 14 s	~2 min	10 min	$0.08
Tolstoy chapter (MP3, Spanish)	21 min 54 s	~5 min	22 min	$0.176
Melville's "Bartleby" (MP3, English)	73 min	20.5 min	74 min	$0.592

Speech is transcribed at roughly 0.1-0.3x real time at the default 2 GB memory. Long files are processed in ~10-minute parts cut at silent moments, so a 3-hour file needs no more memory than a 15-minute one. More memory makes a run start faster but does not change the price per minute.

Use it from code, AI agents, n8n and Make

Python (apify-client):

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("kantolabs/audio-video-to-text").call(run_input={
    "fileUrls": ["https://archive.org/download/gettysburg_johng_librivox/gettysburg_address_64kb.mp3"],
    "outputFormats": ["srt", "txt"],
})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["language"], item["text"][:200], item.get("srtUrl"))

One HTTP call that waits for the result:

curl -X POST "https://api.apify.com/v2/acts/kantolabs~audio-video-to-text/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"fileUrls": ["https://archive.org/download/gettysburg_johng_librivox/gettysburg_address_64kb.mp3"]}'

• AI agents (MCP): add the Apify MCP server (https://mcp.apify.com) to Claude, ChatGPT, Cursor or any MCP client; the agent can find this Actor and call it whenever it needs to read what is said in an audio or video file.
• n8n: use the Apify node, operation Run actor and get dataset, actor kantolabs/audio-video-to-text, input {"fileUrls": ["{{ $json.fileUrl }}"]}, then map text or srtUrl into the next node.
• Make / Zapier: Apify module Run an Actor (wait for finish) -> Get dataset items.
• Schedules and webhooks: transcribe new uploads every hour, or get a webhook when a run finishes.

Limits

• Direct file links only. YouTube, TikTok, Instagram, Facebook, X, LinkedIn, Vimeo, Twitch, Douyin, SoundCloud and Spotify pages are refused on purpose - their terms forbid downloading.
• Google Drive and Dropbox files must be shared as "anyone with the link". Private S3/R2/GCS objects need a pre-signed URL.
• Up to 300 minutes and 4 GB per file.
• No speaker labels (diarization) and no translation yet - the transcript does not say who speaks.
• Whisper base can mishear names and rare terms, and is weak on some languages: in our tests Hindi came out in Urdu script and Japanese had many kanji errors. For those, check the result on a short sample first. If the model gets lost on very unclear audio, the Actor stops there and charges only the minutes it transcribed.
• Music-only and silent files return a free "no speech" row; sung lyrics are transcribed poorly.

FAQ

Which languages are supported? The 99 Whisper languages, including English, Spanish, French, German, Portuguese, Italian, Dutch, Polish, Russian, Ukrainian, Turkish, Arabic, Hindi, Japanese, Chinese, Korean, Indonesian and Vietnamese. Leave language on auto to detect it per file.

Which model is it? OpenAI Whisper base (74M parameters), int8, with Silero voice-activity detection. It is fast and cheap and does well on clear speech in English, Spanish, French, German, Italian, Portuguese, Dutch and other widely spoken languages. Transcripts include punctuation and capitalisation.

Can I get subtitles for my video? Yes - every successful file gets an SRT and a VTT file with cues of at most two 42-character lines, timed to the speech.

Is it the real OpenAI Whisper? Yes - the open-source Whisper base model (MIT licence) run with faster-whisper (int8 CTranslate2) on Apify servers. Your files are not sent to OpenAI or any other API.

Does it work with YouTube links? No. Download rights belong to the platform and the creator; this Actor only processes files you can link to directly.

What happens to my files? They are downloaded to the run's temporary disk, transcribed and deleted when the run ends. The transcripts stay in your run's storage under your Apify account.

Something is not working? Open an issue on the Issues tab with the file link - it is answered quickly.

Related Actors by Kanto Labs

• Video to LLM Context - transcript aligned to each scene plus a keyframe per scene, ready for AI agents
• Auto Captions & Lyric Video Maker - burn TikTok-style captions into a video or make a karaoke lyric video
• Video Frame Extractor - keyframes, scene changes and thumbnails from any video, ready for AI vision
• Image to Text OCR - text from photos, screenshots and scanned PDFs
• Document to Markdown - PDF, Word, PowerPoint and Excel to clean Markdown for LLMs