LAB QUOTA · OK
[ podcast-clips:// ] experimental
cat: audio model: @cf/meta/llama-3.1-8b-instruct

Long-form audio transcript → 5 suggested 60-second clip windows with timestamps + the "why this moment".

// system prompt
You suggest clip windows from long-form audio transcripts. User pastes transcript + audio type. Output 5 clips:

  ## Clip N — <one-line title>
  Timestamps: <HH:MM:SS – HH:MM:SS> (~60s)
  Excerpt:
  > <transcript of the clip window>

  Why this works: <one-line — what makes this standalone>
  Suggested caption: <social-post caption, under 200 chars>

Rules:
- Target ~60s windows (45-75s acceptable). They're the right length for Reels / TikTok / LinkedIn video.
- Each clip must work STANDALONE — listener doesn't need the rest of the audio to get it.
- Look for: contrarian takes, concrete examples, surprising data, vulnerable moments, sharp 1-liners with setup.
- Avoid: clips that depend on inside jokes, ones where the most interesting part is the response and you're cutting before it.
- Suggested caption is platform-flexible (works as a tweet, LinkedIn opener, or Reels overlay).
⚡ powered by Cloudflare Workers AI · quota deducted on success
// sample output
## Clip 1 — "The 100-passionate-users vs 10,000-shruggers" framework
Timestamps: 00:12:34 – 00:13:35 (~61s)
Excerpt:
> "When we were debating MAU vs activation metrics, our CEO said something that's stuck with me. He said, 'I'd rather have 100 customers who would be devastated if we shut down than 10,000 who shrug.' That single sentence changed how we thought about retention. We stopped optimising for top-of-funnel and started measuring 'who would be sad if we disappeared?' That's the metric."

Why this works: A sharp idea, with the setup, the quote, and the takeaway in 60 seconds. The "100 vs 10,000" framing is repeatable — viewers will quote it themselves.
Suggested caption: "100 customers who'd be devastated > 10,000 who'd shrug. The metric most product teams aren't measuring."

## Clip 2 — "Most roadmaps die in steering committees"
Timestamps: 00:24:11 – 00:25:05 (~54s)
Excerpt:
> "Most product roadmaps die in steering committees, not in the build. The build was fine. The build was always fine. What kills them is six weeks of 'let's align with the regional VPs' and 'we should run this past compliance' and by the time you start writing code, your six-month plan has become a nine-month plan that everyone in the room is exhausted by. The fix is bandwidth, not better planning."

Why this works: Contrarian + named cause. Names the enemy (process), not the people. Pairs well with the typical product-management audience.
Suggested caption: "Most roadmaps die in steering committees, not in the build. The build was fine. The build is always fine."

## Clip 3 — "Shipping slower made the team better"
Timestamps: 00:39:48 – 00:40:52 (~64s)
Excerpt:
> "Here's the counterintuitive thing — every time we slowed down our shipping cadence, the team felt better about the work. We went from weekly releases to bi-weekly, and on paper that's a 50% slowdown. But the work that shipped was better, the team was happier, the customer feedback was sharper. Sometimes the fastest way to improve velocity is to ship less, more deliberately. That's a hard sell at a startup. But it worked."

Why this works: Counterintuitive claim with a specific example. Sets up the next "and here's why" conversation.
Suggested caption: "Shipping slower made the team better. Same hours, better work. Sometimes velocity isn't the answer."

## Clip 4 — "Stop trying to sell to everyone"
Timestamps: 00:48:22 – 00:49:18 (~56s)
Excerpt:
> "Our growth came from one thing — and I've told this story a few times now — we stopped trying to sell to everyone. We picked one customer profile, narrowed our messaging to them, and turned away every lead that didn't fit. We lost 30% of our pipeline in the first quarter. Six months later we were growing twice as fast and our churn was half. The math wasn't about funnel size, it was about who was in the funnel."

Why this works: Lesson with a specific number, a specific tactic, and the result. Replicable.
Suggested caption: "We lost 30% of our pipeline by saying no to non-ICP leads. Six months later, growth was 2x. The math isn't funnel size — it's who's in it."

## Clip 5 — "Active doing what?"
Timestamps: 01:02:15 – 01:03:08 (~53s)
Excerpt:
> "If your retention metric is just 'monthly active', you're lying to yourself. Active doing what? Logging in? Browsing the dashboard? Closing the tab after 8 seconds? That's not retention, that's checking the locked door is still locked. Real retention is 'doing the thing you came here to do, repeatedly'. Find the verb that defines your product and count that. 'Monthly active' is the easiest metric to game and the hardest one to learn from."

Why this works: Strong opener question ("active doing what?") makes a great video hook. Specific tactical advice (find the verb).
Suggested caption: "If your retention is just 'monthly active' — active doing what? Find the verb your product depends on. Count THAT."
// powered by cloudflare workers ai · quota deducted on success ← back to catalog