Turn a podcast episode into show notes in 20 minutes
A practical recipe for taking a 60-minute recorded podcast episode through transcription to ready-to-publish show notes, chapters, and social cuts in roughly 20 minutes of working time.
Show notes are the part of podcast production that nobody enjoys and everybody needs. The episode is recorded, the audio is edited, and the publishing window is closing while you stare at a blank document trying to remember what your guest said in minute 34. The traditional answer is "spend two hours writing them by hand." The 2026 answer is "transcribe the episode, then write the notes from the transcript in twenty minutes."
This is the recipe.
音频转文字
将音视频转为文本，支持字幕导出
What you actually need
You need three artifacts at the end:
- Episode description (~150-300 words). The thing that goes in the podcast platform's episode field, gets emailed to subscribers, and appears as the meta description on the episode page.
- Show notes proper (~600-1500 words). The longer-form summary, with chapters, timestamps, links, guest credentials, and quotes. This is what serious listeners read before deciding whether to subscribe.
- Social cuts (3-6 short quotes or one-liners). The bits you post on X, LinkedIn, Threads, or Bluesky to drive traffic.
All three are derivable from a transcript with timestamps. You do not need to listen to the episode again. You barely need to skim it.
The recipe
Step one (5 minutes): transcribe the episode.
Generate detailed JSON output. Enable speaker diarization with min and max set to the actual number of speakers. Enable word-level timestamps if you want pull-quotes with precise timing. Add a prompt that lists your show name, guest names, and any unusual terms or proper nouns the conversation covered. The prompt is the difference between "John Smith" being captured correctly and "John Smith" appearing as "Jon Smyth" in twelve places.
For a 60-minute episode at typical bitrates, transcription itself takes 1-3 minutes of wall time depending on the service. Submit it, then go make coffee.
Step two (5 minutes): scan and pull chapter markers.
Open the transcript. Skim it (do not read every word). Look for the natural seams in the conversation: where the topic shifts, where you wrap a segment, where the guest answers your big question. Note the timestamp of each one.
For most podcasts, you end up with 5-10 chapters. Write them down as a list with their start times. This is your chapter-marker file (most podcast players support these, including Apple Podcasts, Pocket Casts, and Overcast).
Step three (7 minutes): draft show notes from the transcript.
Copy the transcript into your notes draft. Now do three passes.
First pass: delete everything that is not load-bearing. Transitions, throat-clearing, "and so, you know, the thing about that is" filler, repeated false starts, and any stretch of the conversation that did not produce a memorable point. The transcript shrinks to maybe 30 percent of its original size.
Second pass: turn the remaining content into proper paragraphs. Rewrite the spoken-language fragments into readable prose. This is where you might add a sentence to set context, restate a point more clearly than the speaker did, or smooth a transition.
Third pass: add structure. Pull out the chapter markers from step two as ### Heading-style sections. Add timestamps where a specific quote or moment is worth linking to. Hyperlink any external references the guest mentioned.
The result is the long-form show notes. It will feel like writing a blog post, because that is what it is.
Step four (3 minutes): pull the social cuts.
Go back to the transcript with chapters in mind. Pick 3-6 quotes that work standalone, paste each into a list, attach the speaker and the timestamp. These are your social posts and email teasers.
Useful criteria for what makes a good cut: it is short (under 280 characters works on every platform), it has a hook (a strong claim, a counterintuitive observation, a specific number), and it does not require knowing the rest of the conversation to land. "We tested 200 prompts and the only thing that mattered was specificity" works as a cold tweet. "And so as I was saying" does not.
What the transcript fixes that you used to do by hand
The transcript collapses three traditionally manual tasks: relistening to find quotes (the timestamps are right there), trying to remember what was said in minute 34 (you can read it), and deciding what was important (you can scan the whole conversation in 90 seconds instead of 60 minutes).
The transcript does not write the show notes for you. The drafting pass is still your job, because the transcript captures every word said, and the show notes need to capture only the words worth saying again. That distinction is editorial, and editorial is a human task.
Cost math, briefly
A 60-minute episode transcribed at typical per-minute pricing is well under a dollar. A two-hour episode is about a dollar to two dollars. For a podcast that publishes weekly, the transcription cost over a year is less than the cost of a single freelance editor's hour. The savings are not really in money; they are in calendar time and creative energy.
Optional add-ons
Once the workflow is set up, two cheap additions are worth knowing about:
An RSS audio search index. Take the timestamped transcripts of your full archive, index them in any text-search system (Algolia, Meilisearch, Postgres full-text), and add a search box to your podcast site that lets listeners find the moment a specific topic was discussed. The combined cost is a few hours of one-time setup and the per-minute transcription on episodes that have not been transcribed yet.
A guest-quote highlight reel. If you publish guest interviews, a per-episode pull-quote graphic for social is easier when the transcript exists than when it does not. You already have the quotes from step four; running them through a quote-card design template is a five-minute exercise.
The honest caveat
The 20 minutes assumes a 60-minute episode and a workflow you have used before. Your first time through, plan for an hour: half of that is figuring out which prompt your show needs and how aggressive your editorial trim is. By the third episode, you are at 20 minutes. By the tenth, you are checking your watch wondering if you missed a step.
If you publish weekly, this workflow saves you roughly an hour per episode versus writing notes from memory. Over a year of weekly episodes, that is a working week of time recovered. For a hobby show, that is the difference between maintaining the show and quietly letting it die.
Audio Transcription · Z.Tools
Convert audio and video to text with subtitles
