workflow · voice
Blog post → podcast in 20 minutes with Claude + ElevenLabs
Verdict: The highest-ROI content repurposing workflow we run. One written piece becomes two channels for ~$0.30 of API cost and 20 minutes.
Heads up: this post contains affiliate links. If you buy through them, we may earn a commission at no extra cost to you, it's how the site stays free. We only recommend tools we've actually run ourselves. Full disclosure.
I ship AI-narrated audio that customers pay for, so this isn’t a theory post. Here’s the exact pipeline, including the part everyone skips, why you can’t just paste a blog post into a voice tool.
Why pasting fails
Written English and spoken English are different languages. A blog post read aloud sounds like a blog post read aloud: subheads become awkward pauses, parentheticals die, lists sound robotic. The fix is a spoken-word rewrite pass, and it’s what Claude is for.
Step 1: The rewrite prompt
Rewrite this article as a spoken monologue for audio narration. Rules: contractions everywhere; no subheadings, use spoken transitions (“here’s the thing”, “so what does this mean?”); break sentences over 20 words; convert lists into flowing speech; add one rhetorical question per section; open with a hook, not the title; close with a single takeaway sentence. Keep every fact. Target 90% of original length.
Paste your post under that prompt. Claude returns a script that sounds like talking.
Step 2: Voice settings that don’t sound AI
In ElevenLabs, the defaults are tuned for demos, not narration. What we ship with: stability around 45–55 (lower = more expressive, higher = more monotone), similarity 75+, and, the one everyone misses, generate section by section, not the whole script at once. Long single generations drift flat; sectioned generations stay energetic, and regenerating one flat paragraph beats regenerating twelve minutes.
Step 3: Assembly
Stitch sections, normalize loudness, top-and-tail with 2 seconds of room tone. Any free editor does this, the tools aren’t the constraint. Total marginal cost per episode on the Creator plan: cents.
What this unlocks
Every evergreen post becomes a podcast episode. Your PDF product grows an audio tier at zero marginal cost. Your course gets narration without a microphone. One written asset, three formats.
The full prompt chain, rewrite pass, tone-matching, and the QA checklist we run before shipping customer audio, is in the prompt pack:
📄 Free: The Audio Pipeline Prompt Pack
The spoken-word rewrite, tone matcher, section splitter, and shipping QA checklist, ready to paste into Claude.