Podcast transcription: Transcripts, chapters, show notes and quotes

Start from the published file

Transcribe the edited episode you will publish, not the raw recordings. Every minute cut in editing shifts the times, and chapters taken from the raw files land in the wrong place in the episode.

1. On New transcript, choose the file. MP3, WAV, M4A and video files are taken as they are.
2. Choose the language, and leave Tell speakers apart on.
3. Choose in the background: the episode is written without being played, as fast as the device can keep up.

You can pick a range to skip the theme music. Times always count from the start of the file, so a range does not shift your chapter times. Over music the model sometimes writes words nobody said: glance at the text at the start and the end.

Naming hosts and guests

Speakers first appear as “Speaker 1”, “Speaker 2”. Name one and the name changes across the whole transcript. If one person was split into two numbers, give both the same name and they merge. Check the labels where two hosts sound alike or everyone laughs at once. Naming speakers has the details.

Names carry into exported files. In VTT the line where the speaker changes starts with a <v Name> tag; in SRT the name is written at the start of that line as “Name:”.

Chapters from timestamps

In a long episode listeners want to jump to the part they care about. Each paragraph in the transcript shows its time, so finding where the topic changes is enough:

00:00 Intro
02:15 Meet our guest
11:40 How the community gardens started
27:05 The first meeting with the council
41:30 Listener questions

The list goes in the episode description. The platforms’ rules are close:

• Spotify: At least three chapters, the first at 00:00, times in order, titles in plain text without emoji. See Spotify’s chapters page.
• YouTube: The first time 00:00, at least three times in ascending order, each chapter at least 10 seconds long. See YouTube Help.

The same list works on both, so write it once.

Writing show notes

Show notes are the short text people read before they listen. The transcript helps most in finding what was mentioned and nobody wrote down: books, sites, people. Search the text for “book”, “website”, “link”. An example:

Episode 12: Community gardens

We talked to Elif Demir, who set up seven community gardens in three years, about turning an empty lot into a garden, the first meeting with the council, and why the neighbours kept their distance in the first year.

Mentioned: “Soil in the City”, the book Elif recommends; the gardens’ shared calendar; Murat, who sent in the listener question.

Chapters: The list above.

The episode and the book are made up for the example. Do not copy the first sentence of your notes from the transcript; write it yourself. Spoken sentences read long and loose.

From transcript to blog post

Posting the raw transcript as an article does readers no favours: speech repeats itself, sentences break off and topics wander. A better route:

1. Pick the episode’s three main ideas and give each a subheading.
2. Under each, quote the best sentences from where that topic came up, and link them with your own words.
3. Put the full transcript below the post or on its own page, where a listener looking for a line can find it.

You can add headings, bold and lists in the transcript, then download it as Markdown or Word. Markdown pastes straight into many blogging tools.

Picking a quote and checking it against the audio

A shared quote must be exactly what the guest said. A good one stands alone, finishes a thought and stays short:

• “Soil doesn’t make a garden, neighbours do.” Stands alone.
• “So, like, that’s why it ended up that way anyway.” Meaningless out of context.

Before you publish a quote, click its paragraph time and listen. The model may have misheard a word, and putting words in a guest’s mouth costs a podcast its credibility. If you shorten it, mark the gap with “[…]”. The card itself comes from your design tool: LiveType gives you the text, not the image.

Which podcasts, and the limits

• This guide is for your own podcast or episodes you hold the rights to. Downloading someone else’s episode to republish it infringes their copyright.
• LiveType does not write summaries, chapters or quote cards, and does not translate: it writes what is said and you do the rest.
• Crosstalk, laughter and background music weaken the text.
• When you transcribe on your device, the audio does not leave your computer, which matters for unreleased episodes (private transcription).

The free plan has 30 minutes in total, enough to try an episode. For a weekly show, see pricing. Subtitle files are covered in more depth in our subtitle generator guide.