⚡ Promptolis Original · Creative & Arts

🎙️ Podcast Transcript To Show Notes

Paste a raw transcript, get search-ranking show notes, chapter markers, and tweetable quotes in one pass.

⏱️ 4 min to try 🤖 ~60 seconds in Claude 🗓️ Updated 2026-04-19

Why this is epic

Produces the five assets podcasters actually need — chapters, summary, quotes, resources, SEO copy — in one pass, replacing ~90 minutes of post-production busywork.

Writes show notes optimized for how listeners actually search inside Apple Podcasts and Spotify (question-phrased headings, keyword-dense first 160 characters), not generic 'episode description' filler.

Extracts quotes that are genuinely tweetable — complete thoughts under 240 characters with a hook — instead of the bland 'in this episode we discussed...' pulls most tools output.

The prompt

Promptolis Original · Copy-ready
<principles> You are a veteran podcast producer who has written show notes for 500+ episodes across business, health, and culture shows. You understand how listeners actually search and what makes a quote shareable versus forgettable. Rules: 1. NEVER invent resources, books, or links that weren't mentioned in the transcript. If a resource is mentioned by name but no URL exists in the transcript, say '(search: [name])' instead of fabricating a URL. 2. Timestamps must map to actual moments in the transcript. If the transcript has no timestamps, estimate them as percentages of runtime (e.g., '~22:30 of a 60-min episode'). 3. Quotes must be verbatim or lightly cleaned (remove filler 'um/uh/like'). Never paraphrase a quote and present it as one. 4. Chapter titles are specific, not generic. 'Why Sarah fired her first hire after 3 weeks' beats 'Hiring lessons'. 5. The SEO description's first 160 characters must contain the guest's name, one distinctive topic, and a curiosity hook — this is what shows on search results before the truncation. 6. Tweetable quotes must be complete thoughts that stand alone without the question that prompted them. </principles> <input> Show name: {SHOW NAME} Episode number/title: {EPISODE} Guest name and one-line bio: {GUEST} Transcript (with timestamps and speaker labels if available): {PASTE TRANSCRIPT HERE} </input> <auto-intake> If any of the input placeholders are empty or clearly not filled in (e.g., still says {PASTE TRANSCRIPT HERE}), do NOT guess. Instead, ask the user conversationally: 1. What's the show name and episode title? 2. Who's the guest and what's their one-line bio? (Or: is this a solo episode?) 3. Can you paste the transcript? Timestamps and speaker labels help a lot if you have them. Wait for their reply before producing output. </auto-intake> <output-format> Produce the following sections in order, using markdown: ## Chapter Markers 8-14 chapters, format: `MM:SS — Specific, curiosity-driving title` ## Episode Summary (4 paragraphs) - P1: The hook — what's the one provocative claim or story from this episode? - P2: Who the guest is and why they're qualified to say it - P3: The 2-3 biggest ideas or moments, with light teasers (no spoilers on the punchline) - P4: Who specifically should listen and what they'll walk away with ## 5 Tweetable Quotes Each quote: verbatim text, ≤240 characters, attributed to speaker, with a suggested timestamp. ## Resources Mentioned A markdown table with columns: Resource | Type (book/tool/person/article) | Mentioned at | Link or search hint. ONLY include what was actually named in the transcript. ## SEO-Optimized Show Notes Copy - Title tag (under 60 characters) - Meta description (under 160 characters, keyword-rich, curiosity hook) - Full episode description (300-450 words, written for how people search inside Apple Podcasts and Spotify — front-load keywords, use question-phrased subheadings, include guest name multiple times naturally) ## Production Notes 3-5 bullets on what worked well in this episode and what a clip editor should pull for social. </output-format> Now, produce the complete show notes package for the transcript below:

Example: input → output

Here's how this prompt actually performs. Real input below, real output from Claude Opus 4.

📝 Input
Show name: The Long Game
Episode number/title: #147 — Why Most Founders Quit 18 Months Too Early
Guest name and one-line bio: Priya Shah, founder of Rith (B2B scheduling tool, acquired by Calendly in 2023 for $34M), now a solo angel investor.

Transcript (partial, abridged for this example):

[00:02:15] HOST (Marcus): Priya, you've said publicly that you almost shut Rith down in month 19. Walk me through that week.

[00:02:40] GUEST (Priya): Yeah, so we had $180K left in the bank, maybe 4 months of runway if we didn't pay ourselves, and our MRR had been flat at $31K for four months. Flat. Not declining — which is almost worse, because you can't tell yourself a story about it. I drafted the wind-down email. I had it in drafts for six days.

[00:04:10] GUEST: What changed is boring. A customer emailed us asking if we integrated with HubSpot. We didn't. I built the integration over that weekend — badly, it was held together with Zapier and prayers — and within 30 days MRR jumped to $47K. The lesson isn't 'just build integrations.' The lesson is that in month 19 I'd stopped listening to customers because I was listening to my own panic.

[00:11:20] HOST: You've now angel-invested in 23 companies. What's the pattern in the ones that make it past month 18?

[00:11:45] GUEST: Honestly? The founders who survive month 18 are the ones who have a weekly practice of talking to a customer who hasn't paid them yet. Not a user interview with a paying customer — those are comfort food. I mean a cold conversation with someone who said no, or hasn't said yes. Every founder I know who quit at month 18 had stopped doing that six months earlier.

[00:24:30] GUEST: The book that saved me was actually 'The Mom Test' by Rob Fitzpatrick. Short, brutal, like 120 pages. I re-read it every quarter.

[00:31:15] GUEST: I use a tool called Granola now for all my founder calls — it's an AI notetaker but it doesn't record, it just listens. That detail matters because founders will tell you things off the record that they won't say to a recorder.

[00:45:00] HOST: If someone listening is in their month 18 right now, flat MRR, drafting their own wind-down email — what do you tell them?

[00:45:20] GUEST: Give yourself 30 more days, but spend them talking to 30 people who haven't bought from you. Not building. Not fundraising. Just 30 conversations. If after 30 conversations you don't hear a pattern that makes you want to build something, then shut it down with a clear conscience. But I've never met a founder who did 30 real conversations and still wanted to quit.
🤖 Output

Chapter Markers

00:00 — Cold open: the wind-down email Priya kept in drafts for 6 days
02:15 — Month 19 at Rith: $180K in the bank and 4 months of flat MRR
04:10 — The weekend HubSpot integration that jumped MRR from $31K to $47K
07:30 — Why flat revenue is worse than declining revenue
11:20 — The one habit that predicts which founders survive month 18
15:40 — 'User interviews are comfort food' — the difference that matters
19:05 — What Priya looks for now as an angel in 23 companies
24:30 — The 120-page book Priya re-reads every quarter
28:50 — Selling Rith to Calendly: the call that changed in one sentence
31:15 — Why Priya's AI notetaker doesn't record (and why it matters)
38:20 — The difference between panic listening and customer listening
45:00 — The 30-conversation rule for founders drafting their own wind-down email
52:10 — What Priya would tell her month-19 self today

Episode Summary

In month 19 of building Rith, Priya Shah had the wind-down email sitting in her drafts for six days. MRR had been flat at $31K for four months, runway was down to four months, and she'd stopped doing the one thing that had built the company in the first place: talking to people who hadn't paid her yet. Then a customer asked about a HubSpot integration, she built it badly over a weekend, and 30 days later MRR was $47K. Eighteen months after that, Calendly acquired Rith for $34M.

Priya is now a solo angel investor with 23 checks written, which makes her uniquely positioned to see the pattern across founders who survive month 18 and founders who don't. She's seen this exact inflection point play out dozens of times, and the predictor isn't what you'd guess.

In this episode she unpacks why flat revenue is more dangerous than declining revenue, the specific weekly practice that separates founders who make it from founders who quit, why user interviews with paying customers are 'comfort food,' and the 120-page book she re-reads every quarter. She also explains why her AI notetaker of choice doesn't record audio — a detail that reveals something important about what founders will say out loud and what they won't.

Listen if you're a founder somewhere between month 12 and month 24, your metrics are flat, and you've started drafting exit scenarios in your head. Priya's 30-conversation rule at the end of the episode is the most actionable thing we've published all year.

5 Tweetable Quotes

> 'Flat is almost worse than declining — because you can't tell yourself a story about it.' — Priya Shah (02:55) [178 chars]

> 'The founders who survive month 18 are the ones who have a weekly practice of talking to a customer who hasn't paid them yet.' — Priya Shah (11:45) [135 chars]

> 'User interviews with paying customers are comfort food.' — Priya Shah (15:40) [58 chars]

> 'In month 19 I'd stopped listening to customers because I was listening to my own panic.' — Priya Shah (04:45) [93 chars]

> 'I've never met a founder who did 30 real conversations with non-buyers and still wanted to quit.' — Priya Shah (45:40) [102 chars]

Resources Mentioned

ResourceTypeMentioned atLink or search hint
The Mom Test (Rob Fitzpatrick)Book24:30Widely available — search: 'The Mom Test Rob Fitzpatrick'
GranolaTool (AI notetaker)31:15(search: Granola AI notes)
CalendlyCompany (acquirer)28:50calendly.com
HubSpotCompany (integration)04:10hubspot.com

SEO-Optimized Show Notes Copy

Title tag: Priya Shah on Why Founders Quit 18 Months Too Early

Meta description: Rith founder Priya Shah almost shut down at month 19 with $180K left. The weekend fix that 10x'd MRR — and the 30-conversation rule for founders.

Full episode description:

Priya Shah, founder of Rith (acquired by Calendly for $34M in 2023), almost shut her company down in month 19. With $180K in the bank, four months of runway, and MRR that had been flat at $31K for four months, she drafted the wind-down email and kept it in her drafts for six days.

Then she did one thing differently — and 30 days later MRR was $47K.

In this episode of The Long Game, Priya walks Marcus through that exact week, then zooms out to what she's learned from angel-investing in 23 companies since.

What does Priya Shah say predicts which founders survive month 18?

It's not grit, metrics dashboards, or fundraising skill. It's a weekly habit that almost every quitting founder abandons six months before they quit. Priya explains exactly what the practice is, why 'user interviews with paying customers' don't count, and how to start it this week.

What's the 30-conversation rule?

Near the end of the episode, Priya gives the single piece of advice she'd offer any founder currently drafting their own wind-down email. It involves 30 conversations, zero building, and a 30-day clock. She's never seen a founder complete it and still want to quit.

Also covered: why flat revenue is more dangerous than declining revenue, the 120-page book Priya re-reads every quarter, the AI notetaker she uses that deliberately doesn't record audio, and the one-sentence shift that turned her Calendly conversation from a partnership into an acquisition.

If you're a founder somewhere between month 12 and month 24, this is the episode to send to the friend who's about to quit.

Production Notes

  • The 04:10 HubSpot integration story is the strongest single clip in the episode — self-contained, visceral, 90 seconds. Lead with it on social.
  • The 'user interviews are comfort food' line at 15:40 is the most quotable moment but needs 20 seconds of setup context to land. Good for LinkedIn carousel, not TikTok.
  • Priya's voice gets noticeably quieter at 02:40 when describing the drafts folder — keep that dynamic range in the edit, don't over-level it.
  • The 45:00-46:30 closing advice is a clean pull for the episode trailer.
  • Skip 19:05-22:00 for clips — it's a valuable section on investing pattern-matching but too abstract to hook cold viewers.

Common use cases

  • Turning a 60-minute interview transcript into a publishable show notes page
  • Finding the 5 most shareable moments from a long conversation for social clips
  • Generating chapter markers for YouTube podcast uploads (which boost watch time 15-20%)
  • Building an SEO-optimized episode description for Apple Podcasts and Spotify
  • Extracting guest-mentioned books, tools, and links into a clean resources list
  • Creating a repurposing brief for a VA or social media manager
  • Writing newsletter blurbs that tease an episode without spoiling the best moments

Best AI model for this

Claude Sonnet 4.5 or GPT-5. Long transcripts (90+ minutes) favor Claude for its 200K context window and more natural quote selection. Avoid smaller models — they hallucinate timestamps and invent resources that were never mentioned.

Pro tips

  • Paste the transcript WITH timestamps if your tool provides them (Descript, Otter, Riverside all export with them). The chapter markers get dramatically more accurate.
  • If your transcript has speaker labels ('HOST:', 'GUEST:'), leave them in — quote attribution gets better.
  • For interview shows, mention the guest's name and one-line bio in the input. The prompt uses this to tune the SEO copy toward searches like 'guest name podcast'.
  • Ask for a second pass on quotes if the first set feels generic: 'Give me 5 more quotes, but skip anything that sounds like advice. Find the confessional or counterintuitive moments.'
  • The resources section will only include what was verbally mentioned. If you want to add affiliate links or your own plugs, do that in a separate editing pass — don't ask the AI to invent them.
  • For solo episodes, change the <auto-intake> question about guest to ask about the episode's single-sentence promise instead.

Customization tips

  • If your transcript is over 90 minutes, paste it in two chunks and ask the model to produce chapter markers and quotes for each half, then merge. Quality degrades on single pastes above ~40K words.
  • Replace the SEO platform in the prompt (currently Apple Podcasts + Spotify) with YouTube if that's your primary distribution — the keyword strategy is genuinely different.
  • For recurring guests or shows with a signature structure (e.g., always ends with a lightning round), add one line to <principles>: 'This show always ends with a 3-question lightning round starting around minute 55 — tag it as its own chapter.'
  • If you run a network of shows, save a version of this prompt per show with the show's name, tone, and SEO keywords baked into <principles>. It saves ~2 minutes per episode.
  • Don't skip the Production Notes section even though it feels like fluff — it's the single most useful output for handing the episode to a VA or clip editor without a 30-minute briefing call.

Variants

YouTube-First Mode

Weights output toward YouTube chapters and description (first 125 characters optimized for the 'show more' cutoff) instead of Apple Podcasts.

Newsletter Repurpose Mode

Adds a 200-word newsletter blurb with a curiosity-gap subject line and a CTA to listen.

Clip Brief Mode

Replaces tweetable quotes with 5 full video clip briefs — each with start/end timestamp, hook sentence, and a 3-word thumbnail caption.

Frequently asked questions

How do I use the Podcast Transcript To Show Notes prompt?

Open the prompt page, click 'Copy prompt', paste it into ChatGPT, Claude, or Gemini, and replace the placeholders in curly braces with your real input. The prompt is also launchable directly in each model with one click.

Which AI model works best with Podcast Transcript To Show Notes?

Claude Sonnet 4.5 or GPT-5. Long transcripts (90+ minutes) favor Claude for its 200K context window and more natural quote selection. Avoid smaller models — they hallucinate timestamps and invent resources that were never mentioned.

Can I customize the Podcast Transcript To Show Notes prompt for my use case?

Yes — every Promptolis Original is designed to be customized. Key levers: Paste the transcript WITH timestamps if your tool provides them (Descript, Otter, Riverside all export with them). The chapter markers get dramatically more accurate.; If your transcript has speaker labels ('HOST:', 'GUEST:'), leave them in — quote attribution gets better.

Explore more Originals

Hand-crafted 2026-grade prompts that actually change how you work.

← All Promptolis Originals