Audio to Text & Podcast Transcriber: Speech to Text, SRT avatar

Audio to Text & Podcast Transcriber: Speech to Text, SRT

Pricing

from $15.00 / 1,000 audio minute (tiny/base)s

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

Audio to Text & Podcast Transcriber: Speech to Text, SRT

Audio to text and speech to text for MP3, video files and podcast RSS feeds: full text, SRT and VTT subtitles with timestamps. Whisper in 99 languages, automatic language detection, translation to English, only new episodes on a schedule. Silent and failed files are free.

Pricing

from $15.00 / 1,000 audio minute (tiny/base)s

Rating

0.0

(0)

Developer

Makerio

Makerio

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

2 minutes ago

Last modified

Categories

Features

Limited permissions

Share

What does Audio to Text & Podcast Transcriber do?

Audio to Text & Podcast Transcriber is a speech to text tool: it turns audio and video files and whole podcast feeds into text (audio to text, video to text, MP3 to text). Give it direct links to files (MP3, M4A, WAV, OGG, OPUS, FLAC, AAC, MP4, WEBM, MOV and more) or a podcast RSS feed (an Apple Podcasts link works too: the Actor finds the show's RSS feed or the episode's audio file), and get back:

  • the full transcript in paragraphs,
  • timestamped segments (start and end of every phrase),
  • SRT and VTT subtitle files, a plain TXT file and a Markdown file with timestamped paragraphs (ready for show notes or a blog post) for each recording,
  • the detected language, duration and word count.

Speech recognition runs on Whisper through whisper.cpp, inside the Actor itself: your audio is not sent to any third-party speech API, and you need no API key of your own. 99 languages are supported, with optional translation to English.

Why use it?

  • Podcasters and producers: show notes, blog posts, subtitles and searchable archives of every episode.
  • Researchers and journalists: interviews, lectures and public recordings as text you can search, quote and analyse.
  • Marketers and SEO: turn audio content into indexable pages; monitor what competitors' podcasts talk about.
  • Developers and data teams: feed transcripts to search engines, summarizers, RAG pipelines and language models through the Apify API.
  • Accessibility: captions (SRT/VTT) for video players and course platforms.

Pricing: pay per audio minute

ModelPrice per audio minute
tiny, base$0.012
small$0.04
  • One minute = one started minute of transcribed audio per file, so a 4 min 10 s episode costs 5 minutes.
  • Never charged: files that cannot be downloaded or decoded, and files without speech (silence or music only).
  • Platform usage is included in the price. Set a maximum cost for the run in Apify Console: the Actor stops before it, cutting the last file at the minute the budget covers.

Example: 10 podcast episodes of 45 minutes with the base model cost 450 × $0.012 = $5.40.

How to transcribe a podcast or an audio file

  1. Click Try for free.
  2. Paste links to audio or video files into Audio or video file links, or a podcast feed into Podcast RSS feeds.
  3. Choose the model and, if you know it, the spoken language.
  4. Click Start. When the run finishes, open the Transcripts tab, or download the dataset as JSON, CSV or Excel. SRT, VTT and TXT files are linked from every row.

Input example

{
"audioUrls": ["https://archive.org/download/short_poetry_001_librivox/because_I_could_not_stop_for_death_dickinson_64kb.mp3"],
"rssFeeds": ["https://librivox.org/rss/88"],
"maxEpisodesPerFeed": 3,
"onlyNewEpisodes": false,
"model": "base",
"language": "auto",
"translateToEnglish": false,
"prompt": "",
"skipSilence": true,
"maxMinutesPerFile": 240,
"saveSegments": true
}

Output example

One dataset item per file or episode:

{
"url": "https://archive.org/download/short_poetry_001_librivox/because_I_could_not_stop_for_death_dickinson_64kb.mp3",
"source": "file",
"title": "because_I_could_not_stop_for_death_dickinson_64kb",
"durationSec": 47.3,
"language": "en",
"model": "base",
"task": "transcribe",
"billedMinutes": 1,
"truncated": false,
"noSpeech": false,
"text": "Because I could not stop for death by Emily Dickinson. Because I could not stop for death, he kindly stopped for me...",
"wordCount": 131,
"segmentCount": 14,
"segments": [
{ "start": 0, "end": 4.2, "text": "Because I could not stop for death by Emily Dickinson." }
],
"srtUrl": "https://api.apify.com/v2/key-value-stores/.../records/because-i-could-not-stop-for-death-dickinson-64kb-1a2b3c4d.srt",
"vttUrl": "https://api.apify.com/v2/key-value-stores/.../records/because-i-could-not-stop-for-death-dickinson-64kb-1a2b3c4d.vtt",
"txtUrl": "https://api.apify.com/v2/key-value-stores/.../records/because-i-could-not-stop-for-death-dickinson-64kb-1a2b3c4d.txt",
"processingSec": 9,
"error": null
}

Podcast episodes also have feedUrl, feedTitle, episodeTitle, episodeId and published. A file that fails gives a row with error and errorType and is not charged.

Podcast feeds and "only new episodes"

Add the RSS feed of any podcast. The Actor reads the episode files from the feed (enclosures) and transcribes the newest Episodes per feed of them. Feeds without dates, such as audiobooks, are taken in their own order.

Turn on Only new episodes and schedule the Actor daily or weekly: each run transcribes only episodes it has not transcribed before, so you pay once per episode. The list of done episodes is kept per feed in the key-value store podcast-transcriber-feeds of your Apify account; delete it to start over.

Choosing a model

ModelSpeedAccuracyUse for
tinyfastestroughquick drafts, keyword search in large archives
basefastgood on clear speechpodcasts, audiobooks, lectures, interviews with good microphones
smallabout 5× slower than basebest of the threeaccents, noisy recordings, phone calls, less common languages

Measured on Apify with one CPU core: the base model transcribes an hour of speech in about 20 minutes, tiny in about 7 minutes, small in about 1 hour 40 minutes. Several files run one after another in the same run, and the next file is downloaded while the current one is transcribed.

Tips for better transcripts

  • Set the spoken language when you know it. Automatic detection listens to the first 30 seconds; a music intro can mislead it.
  • Names and terms: list names, brands and jargon in the optional field (for example "Kubernetes, Grafana, Dr. Okafor"). Whisper then spells them as written.
  • Skip silence and music (on by default) detects speech first. It speeds up files with long pauses and prevents invented text in silent parts. Timestamps still match the original file.
  • Translate to English writes the transcript in English whatever language is spoken.
  • Memory: 4 GB gives one full CPU core and is the right setting for most runs. 8 or 16 GB make long files finish sooner; the price per minute stays the same.

Limits

  • Links must be public direct links to files or RSS feeds. Pages of YouTube, Spotify, Apple Podcasts, TikTok, Instagram, SoundCloud and similar sites are not supported; files behind a login or a signed link that expired cannot be downloaded.
  • Up to 2 GB per file. Files longer than Max minutes per file (240 by default, up to 720) are cut, and only the transcribed part is charged.
  • Speaker labels (who said what) are not included.
  • Transcripts are machine-made: check names and numbers before you publish them.

Using the API

Start a run and get the transcripts in one call with the Apify API:

curl -X POST "https://api.apify.com/v2/acts/makerio~podcast-audio-transcriber/run-sync-get-dataset-items?token=YOUR_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{"audioUrls": ["https://example.com/episode-42.mp3"], "model": "base"}'
import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: 'YOUR_API_TOKEN' });
const run = await client.actor('makerio/podcast-audio-transcriber').call({
rssFeeds: ['https://example.com/podcast.rss'],
maxEpisodesPerFeed: 5,
onlyNewEpisodes: true,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const ep of items) console.log(ep.episodeTitle, ep.wordCount, ep.srtUrl);
from apify_client import ApifyClient
client = ApifyClient("YOUR_API_TOKEN")
run = client.actor("makerio/podcast-audio-transcriber").call(run_input={
"audioUrls": ["https://example.com/interview.m4a"],
"language": "es",
"translateToEnglish": True,
})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
print(item["title"], item["text"][:200])

Connect the results to Google Sheets, Slack, Zapier, Make or your webhook with Apify integrations.

FAQ

Which audio formats work? Anything ffmpeg can read: MP3, M4A, M4B, AAC, WAV, OGG, OPUS, FLAC, WMA, AIFF, AMR, and the audio track of MP4, MOV, MKV, WEBM and other video files.

How accurate is it? On clear speech the base model gets most words right; small does better on accents and noise. Accuracy depends on the recording more than anything else.

How long does it take? It depends on the model and the length of the audio; see the model table. Several files are processed one after another, and the next file downloads while the current one is transcribed.

Is my audio stored or shared? Files are downloaded to the run's temporary disk, transcribed there and deleted. Transcripts are saved only in your own dataset and key-value store.

Can it transcribe YouTube videos? No. Use a direct link to an audio or video file, or the podcast's RSS feed.

What does a "minute" mean for billing? Each file is rounded up to whole minutes of transcribed audio. A run with a 30-second clip and a 61-second clip is charged 1 + 2 = 3 minutes.

Something went wrong. Open the Errors view of the dataset: each failed link has a reason. Failed files are never charged. You can also report an issue on the Issues tab.

More tools by Makerio