Make podcast show notes from a transcript

What you get

# Small Kitchen 12: Bread and the market stall

[Write the description here: two or three sentences saying who is speaking and what the episode answers.]

**Speakers (read from the transcript):** Host, Dana

## Chapters
0:00 Oven, loaves, pot
3:03 Pricing, market, charge

## Quotes to check and use
> The biggest mistake was the oven.
> Dana, 0:41

## Links mentioned
- https://example.com/market

## Names and terms to check
Omar (2x)

## Length
8:30 of audio, about 760 words in the transcript.

That is the shape; the words come from your transcript (the example above is made up). Every part can be deleted, and it is plain Markdown or plain text, so it pastes into a podcast host, a blog post or a notes app.

What each part is, and how it is found

• Chapters. One about every 2, 3, 5, 10 or 15 minutes, the first at 0:00. Each starts near its time at the longest pause if your file has end times (SRT, VTT), otherwise after a full stop. The default titles are the three words that are most particular to that stretch (English text only); the other choices are the first words of the line or "Chapter 3". They are placeholders to rename. The rules are on the chapters page.
• Quotes. Whole sentences between 10 and 40 words, scored down for filler words, for starting by pointing back ("and that was why...") and for being a question, and up for words like never, biggest or mistake and for numbers. They are kept a minute apart. The scoring is on the quotes page; it is arithmetic, not an understanding of the episode.
• Links mentioned. Only web addresses that are written out in the transcript (https://...). A speech recogniser usually writes "example dot com" as words, and the tool does not guess at those.
• Names and terms to check. Capitalised words that appear in the middle of a sentence at least twice. It has no dictionary: it will list a place or a brand as readily as a person, and it will miss a name the recogniser wrote in lower case. It is a list of spellings to listen for, since names are what speech recognition most often gets wrong.
• Length. The audio length from the last time in the file, and the word count of the transcript.

All three read the same input: an SRT or VTT file, or text that has a time at the start of each line, such as the timestamped TXT export ([05:12] Maria: ...). Without times there is nothing to place chapters or quotes against, and the page says so. A Name: label at the start of a line is read as the speaker.

What it cannot do

It does not write the description or the title, summarise, find guests' job titles, check that a link works or know which parts are the good parts. Speaker names come from labels in the text; if the transcript has none, the speakers line is left out. A quote is only a candidate until you have listened to it in context, and the times of a sentence inside a long caption are estimated by length. Chapters work best on a transcript with punctuation; a transcript with no full stops falls back to the nearest line.

A workflow that uses it

1. Transcribe the episode and export the transcript with times.
2. If you want a cleaner read, run it through remove filler words first (it keeps the times).
3. Paste it here, set the chapter spacing and copy the draft.
4. Listen to the places where the names and quotes are, fix the spellings in the draft, rename the chapters and write the description. For a write-up, the summary prompts can start one.

For the step before this one, see how to transcribe a podcast and the page for podcast to text; for podcasters in general, podcast transcripts for podcasters.