AI for Business
Podcast Show Notes from a Transcript
Paste or upload a transcript and get the show notes: a summary, chapter timestamps taken from the transcript’s own timings, quotable lines with the time they were said, the links mentioned, the topics, and a short blurb to post. No time is ever estimated and no quotation is ever tidied.
This is a cloud tool. The transcript’s text is sent to the model. It carries everything your guest said, including anything they said off the record — cut those parts out before you paste it. The file itself is read here and never leaves your device. Nothing is stored by us; the call is counted against your month’s allowance — 10 free, 300 on Pro. How the AI tools handle data.
Tips
- Chapters and quote times come from timestamps that are in your transcript. Upload the .srt or .vtt your transcription tool made and they are exact; paste a transcript without times and the tool says so rather than guessing.
- Quotations are copied, not improved. That matters when a guest reads their own show notes — a tidied quotation is a quotation they did not say, and it is the kind of thing that ends a relationship with a guest.
- Where someone says "the link is in the description", the tool records what it was and writes [link needed]. Search for that before you publish; it is usually the one link the episode is for.
- A 45-minute episode is about 7,000 words and reads in one run. An interview over about 90 minutes is better fed in halves — the chapters from each half join up because the times are real.
- The Markdown download drops into most podcast hosts as the episode description, chapter list and all.
Frequently asked questions
What if my transcript has no timestamps?
Then you get everything except chapters, and the notes say plainly that no times were found. The tool is told never to estimate one. Most transcription tools — including the automatic ones in the big editors — will export an .srt or a .vtt with times, and those are the files to use here.
Is my guest’s transcript stored?
No. The text goes to the model through our gateway to be read, the answer comes back to your browser, and neither is kept; only the fact of the call counts against your allowance. If part of the conversation was off the record, cut it out before you paste — that is the only guarantee worth having.
Can it identify who is speaking?
Only as well as your transcript does. If the transcript labels speakers, attribution is right; if it does not, the tool says whose line it could not identify instead of picking a name.
Will it write chapters for YouTube too?
The chapter list is in m:ss and starts at your transcript’s first time. YouTube needs its first chapter at 0:00 and three or more in total; if your transcript starts later than that, add the 0:00 line yourself.