Transcribito

How to transcribe a podcast

From the episode's MP3 to text, and from text to show notes, chapters and subtitles. With tips to keep the result clean.

Why transcribe a podcast

A transcript turns an episode you can only listen to into something you can read, search and quote. It's useful for:

  • Accessibility: people who can't or don't want to listen can read it.
  • Show notes and summaries that don't take you all afternoon.
  • A blog post built from the best of the conversation.
  • Quotes and clips for social media, with the exact minute.
  • Subtitles for the episode's video or for short clips.

How to transcribe your podcast, step by step

  1. Export the audio. An MP3 or WAV from your editor is enough. You don't need maximum quality to transcribe.
  2. Trim the music from the intro and outro if you can. With music or long silences, Whisper can invent sentences; Transcribito discards the most typical ones, but it's best not to give it the chance.
  3. Open the tool, drop the episode and leave the language on automatic detection (or pick the podcast's language).
  4. Choose the accuracy: "Fast" works well with two clear voices in a studio; "More accurate" if there are guests on a video call or noise.
  5. Transcribe, review and download the TXT, or the SRT if you need it for video.

If the episode runs over 3 hours, split it in two. A one-hour episode usually takes a few minutes on a recent computer.

From transcript to publishable content

Clean up the text

Whisper may leave out some filler words, but you can't count on it. Remove the ones that bother you, fix guest and brand names, and add "Host:" and "Guest:" at the start of each turn: the text doesn't tell speakers apart.

Chapters with timestamps

Look for topic changes and note the minute. Platforms such as YouTube recognise a list like "00:00 Intro", "04:12 The first project"; use it in your episode notes too.

Intro and quotes

Pick two or three strong sentences and check each one by playing it from its timestamp before you publish it.

Subtitles for video

If the episode has a video version or you cut clips, use the SRT. For more on how to upload it, see how to create SRT subtitles.

Publishing the transcript

  • Put it as text on the episode page, not only as a downloadable PDF.
  • Say it's an automatic transcript that has been reviewed, and make sure it really has been.
  • You don't have to publish all two hours: a well-edited article with the best parts is often more useful than the raw text.

Unreleased episodes: privacy

Before it goes out, an episode is sensitive material: guest interviews, sponsorships, off-the-record conversations. With Transcribito the file is processed in your browser and isn't sent to a transcription service, so you can prepare the notes without a third party receiving the audio ahead of time.

The Whisper model is downloaded from Hugging Face the first time (and the library from jsDelivr), but that's a download to your device. If you're working with an embargoed episode, also protect the files on your computer and in your backups.

Limits worth knowing

Related guides and uses

Frequently asked questions

Can I transcribe a two-hour episode?

Yes, up to 3 hours per file. On a recent computer it usually takes a few minutes; on a phone it can take far longer, so it's better to do it on a computer.

Does it work if my podcast is in Spanish?

Yes. The language is detected automatically, and you can choose it by hand if you prefer. Whisper works well with English, Spanish and many other languages.

How do I create the episode chapters?

By reading the transcript, noting the minute of each topic change using the timestamps and writing them in the "00:00 Title" format. Transcribito doesn't generate chapters automatically.

Should I publish the full transcript?

You don't have to. For accessibility, offering the full text helps; for a blog, an article built from the best of the episode usually works better.

Does it remove filler words?

Sometimes: Whisper may leave one out, but it isn't reliable at that. Review the text before you publish it.

Does it tell hosts from guests?

No. Whisper doesn't separate speakers, so you'll need to indicate that yourself.