Transcribe a podcast from a link in a few clicks. Paste the episode URL and get a transcript with named speakers, a summary and SRT, PDF or text exports.
To transcribe a podcast from a link, paste the episode's address into VocalStack, press Start, and read the transcript a few minutes later. The link can be an episode page on SoundCloud, Spreaker, iHeartRadio, Audioboom, Mixcloud or YouTube, a podcast's own website, or the episode's MP3. Spotify and Apple Podcasts links need one extra step, covered below.
This guide walks through the whole path in the Dashboard: which links work, how to get the audio address when a link does not, how hosts and guests are named, and how to export the result as text, subtitles or a PDF. A short section at the end does the same through the API.
Most episode links work as they are. VocalStack downloads only the audio, transcribes it, and keeps the link the transcription came from. These are the links to try first:
.mp3, .m4a, .wav, .ogg, .flac or another audio or video format.Two kinds of link are refused before anything is downloaded. A Spotify or Apple Podcasts link is a page for their own app, and VocalStack does not fetch from either. A copy-protected recording cannot be transcribed by anyone else's tool, and the check says so in plain words. Private posts and episodes that need a sign-in are refused too.
The Dashboard does it in five steps.
The transcription is named after the episode's title on its site when the site gives one, so a batch of episodes stays readable in the list. Long episodes have a ceiling: on Pay as you go a file can be up to 2 hours long, and on Premium up to 8 hours. A longer recording is not refused outright. The check tells you its length and lets you set a start and an end, so you transcribe the part you need.
Use the episode's audio file instead of the app link. Almost every podcast is published through an RSS feed, and the feed holds a direct address for each episode's audio. That address works in VocalStack like any other MP3 link.
<item> per episode.<title>, then the <enclosure url="…"> line inside the same item. That URL usually ends in .mp3.Paste the episode's audio address, not the feed address itself: the feed lists every episode, and VocalStack transcribes one recording per link. If the show has a website with a player on each episode page, pasting the episode page is quicker, because VocalStack finds the audio on it for you.
Some shows publish only on Spotify, with no public feed. Those episodes have no audio address outside the app, so there is no link to give VocalStack. If you made the episode, upload your own copy at Transcriptions → New → Upload file instead.
Speaker labels tell the voices apart, and names make them readable. The setting has three values:
A label starts as a number, so give each speaker a name in the transcription. For a show with a regular host, tick Save to Speakers when you name them. The host is then kept on your Speakers page and recognised by voice in later recordings, so next week's episode labels them by name without anyone touching it. Saving a speaker keeps only a voiceprint for recognising them; no voice is cloned unless the person has agreed to it. Telling voices apart is a different job from writing down the words; diarization vs speech recognition explains how the two fit together.
Once the episode is transcribed, the tools for working on it sit beside the transcript.
Every line keeps its time in the episode, so you can play it back and check it before you quote it.
The Export tab turns the transcript into a file, in the spoken language or in any translation:
Through the API, the same export also gives plain text (txt) and Markdown (md), which paste straight into show notes or a blog post.
The Share tab makes a public link to a transcript page where anyone can read along with the audio. You can make several links with different labels, for a guest and for your newsletter, and revoke one at any time; it stops working at once. Readers of a shared link can switch it to their own language and download it.
If the show lives on WordPress, the WordPress transcription plugin transcribes episodes in your Media Library and puts the transcript on the page, where search engines read every word.
A note on other people's shows. Transcribing an episode for your own research, notes or translation is a common use. Publishing someone else's transcript is a different act, governed by copyright where you are and by the show's own terms, so ask the show before you post it.
You can start for free. A new account gets $5 of one-time signup credit, and no card is needed. At the file rate of $0.20 per hour of audio, that credit covers about 25 hours of episodes.
After that, a file costs $0.20 an hour, billed by the second: a one-hour episode is $0.20 and a 30-minute one is $0.10. On Pay as you go every transcription is billed for at least 5 minutes, so a two-minute clip costs the same as a five-minute one. Premium has no minimum and includes $25 of credit every month. Speaker labels, the summary and translations are part of the rate, not extras. The current figures and limits are on the pricing page.
For a backlog of episodes, or a feed you want transcribed every week, the API does the same from code. Create a key at Developers → API keys, then send POST /v1/transcriptions with a JSON body holding the episode's url. It answers at once with the transcription; the server fetches the audio afterwards.
API={{api}}
AUTH="Authorization: Bearer $VOCALSTACK_API_KEY"
TX=$(curl -sSf "$API/v1/transcriptions" -H "$AUTH" -H "Content-Type: application/json" \
-H "Idempotency-Key: $(uuidgen)" \
-d '{"url":"https://example.com/episodes/42.mp3","name":"Episode 42","settings":{"speakerLabels":"auto"}}' \
| jq -r .id)
Read GET /v1/transcriptions/$TX/job every few seconds. It moves from fetching to transcribing to done, or ends in failed with an error that says why. When it is done, export the transcript:
curl -sSf "$API/v1/transcriptions/$TX/export?format=txt" -H "$AUTH" -o episode-42.txt
curl -sSf "$API/v1/transcriptions/$TX/export?format=srt&lang=es" -H "$AUTH" -o episode-42.es.srt
GET /v1/transcriptions/probe?url= checks a link first and tells you whether it opens as a recording, without storing anything. The SDKs wrap all of this as vs.transcriptions.createFromUrl and vs.transcriptions.waitForJob; the file API tutorial walks through them, along with the errors you will meet.
Not from the Spotify link. VocalStack does not fetch from Spotify. Most shows on Spotify also publish an RSS feed, and the episode's audio address from that feed works. Shows that publish only on Spotify have no public audio address, so they cannot be transcribed from a link.
Yes, through its audio file. An Apple Podcasts page is refused, but Apple Podcasts lists shows from their RSS feeds, so the episode's <enclosure> address in the show's feed transcribes like any MP3.
Less than the episode lasts. The audio downloads first, which takes seconds from a podcast host, and transcription runs several times faster than the audio plays. The progress bar shows both stages.
Yes. The spoken language is detected on its own, from 100 languages, and you can name it when you know it, which helps on noisy audio. Any transcript can then be translated into up to 10 languages at the same price.
Yes. Exports, shared links and translations cost nothing beyond the transcription itself, and you can export the same transcript as many times as you like.
Paste your first episode link and read it in a few minutes.