Ai Podcast Creation

Script Generation Framework

Episode Brief (Input for AI)

Before generating any script, define:

Topic: [Specific, not broad]
Audience: [Who is listening, what do they already know]
Episode goal: [Educate / entertain / persuade / update]
Format: [Solo / dialogue / narrative]
Duration target: [5 / 10 / 20 / 45 minutes]
Tone: [Casual / authoritative / storytelling / conversational]
Key points to cover: [3–5 bullet points]
Point of view: [Are we neutral, opinionated, or analytical?]
Call to action: [What should listeners do after?]

AI Prompt for Full Episode Script

Write a [X-minute] podcast script on [topic].

Format: [Solo host / Two hosts named A and B / Narrator-led]
Audience: [Describe]
Tone: [Casual and conversational / Educational and authoritative]

Structure:
1. Cold open (30 seconds): Start mid-thought or with a provocative question
2. Intro (60 seconds): Host intro, episode preview, what listener will get
3. Main content ([X minutes]): Cover these points: [list]
4. Transition bumpers between sections: "Let's talk about..." / "Now here's where it gets interesting..."
5. Outro (60 seconds): Summary, CTA, sign-off

Write in spoken language — contractions, short sentences, natural pauses.
Mark pauses as [PAUSE], emphasis as [EMPHASIS], and section breaks as [MUSIC STING].

Writing for Spoken Audio

Scripts read differently from written text. Follow these rules:

Sentence length: Spoken sentences should be under 20 words. Long sentences cause voice synthesis to lose natural rhythm.

Avoid: Semicolons, em dashes inside sentences, parenthetical asides — these confuse TTS models.

Use instead:

Written: "The results—which surprised even the researchers—showed a 40% improvement."
Spoken: "The results showed a 40% improvement. The researchers were genuinely surprised."

Numbers: Write out numbers in words when they appear mid-sentence.

Avoid: "3 out of 4 founders fail in year 1"
Use: "Three out of four founders fail in their first year"

Technical terms: Spell phonetically in brackets if the TTS model mispronounces:

"The company Xiaomi [shao-mee] announced..."

Multi-Speaker Dialogue Design

For two-host formats, write with distinct voice personalities:

Host A: [Analytical, asks clarifying questions, more formal]
Host B: [Conversational, uses analogies, brings it back to the audience]

Dialogue pacing rules:

• Alternate speaking turns every 2–4 sentences (not longer monologues)
• Include natural reactions: "Right, exactly." / "That's a good point." / "Interesting."
• Write disagreement: authentic podcasts have moments of different perspectives
• Avoid identical sentence structure between hosts

Sample multi-speaker template:

[HOST A]: The report showed that [X]. What's surprising is [Y].

[HOST B]: Yeah, and if you think about it from [audience perspective], that means [implication].

[HOST A]: Exactly. So the question becomes: [question that leads to next point].

[HOST B]: And I think the answer depends on [nuance]. Let me explain what I mean...

Voice Synthesis Tools

Voice Selection Tips

• Avoid choosing the most common default voices — your podcast sounds like everyone else's
• Match voice gender and age to your brand persona
• Test with a 30-second sample before committing to a full episode
• Use "stability" and "clarity" controls (ElevenLabs) to reduce over-emoting on expository content

Intro and Outro Scripts

Intro Template (30–45 seconds)

[MUSIC STING IN]
Welcome to [Podcast Name] — [one-line show description].
I'm [host name/persona].

Today: [3 things the listener will take away, stated as outcomes].

Let's get into it.
[MUSIC FADE]

Outro Template (45–60 seconds)

[MUSIC STING IN]
That's a wrap on today's episode.

The big takeaway: [1-sentence summary of core message].

If this was useful, [subscribe / share / leave a review / follow us].

Next time: [tease next episode topic in 1 sentence].

See you then.
[MUSIC FADE OUT]

Show Notes Generation

Show notes serve SEO and listener convenience. Generate with this prompt:

Write show notes for a podcast episode titled "[title]" about [topic].

Include:
1. Episode summary (3–5 sentences, written for search engines)
2. Key timestamps (format: 00:00 — Topic name)
3. Top 5 takeaways (bullet points)
4. Resources mentioned: [list any if applicable]
5. Guest bio: [if applicable]
6. CTA: subscribe / follow / newsletter / community link

Target length: 300–500 words. Optimize for the keyword: [main topic keyword].

Distribution Checklist

Audio file:

• MP3, 128kbps minimum (192kbps for music-heavy shows)
• Normalized to -16 LUFS (podcast standard loudness)
• Intro and outro trimmed cleanly
• ID3 tags filled: title, artist, album, episode number, cover art

Publishing platforms:

• RSS feed via hosting platform (Buzzsprout, Anchor, Transistor, Podbean)
• Submitted to: Spotify, Apple Podcasts, Google Podcasts, Amazon Music
• Episode title uses searchable keywords (not just episode number)

Supporting content:

• Show notes published to hosting platform and/or website
• Transcript available (accessibility + SEO)
• Audiogram clip (30–60 second video of key quote) for social media
• Episode announced on newsletter and social channels