Audio Script Writing
Parent: Writing and Documentation · Published reference · snapshot 2026-09-08 · skill content-and-marketing-writing/references/audio-script-writing.md
↓ Facts as markdownall context files
Reference for writing that will be heard, not read. The eye can re-scan a sentence; the ear gets one pass.
These notes link each claim to its source. Some citations identify archived research reports that are no longer hosted on this site. A published reference means the content is available; it does not certify independent review or accuracy.Read the editorial policy and follow the sources before relying on a claim.
Audio Script Writing
- Reference for writing that will be heard, not read. The eye can re-scan a sentence; the ear gets one pass. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#audio-script-writing]
The one rule: writing for the ear is single-pass
- Listeners process roughly 150–160 words per minute for narration, 130–145 wpm for voice-UI prompts, and 110–125 wpm for IVR. They lose comprehension on: [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#the-one-rule-writing-for-the-ear-is-single-pass]
- Subordinate clauses stacked more than one deep [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#the-one-rule-writing-for-the-ear-is-single-pass]
- Parentheticals (the ear has no parens) [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#the-one-rule-writing-for-the-ear-is-single-pass]
- Lists longer than three items without explicit numbering [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#the-one-rule-writing-for-the-ear-is-single-pass]
- Numbers and proper nouns delivered without a beat of silence after [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#the-one-rule-writing-for-the-ear-is-single-pass]
Core concept 1 — The one-idea-per-sentence rule
- Eye version (fine on the page): > The migration, which we'd been planning since February, finally launched on Tuesday after a final round of testing that revealed two minor bugs we patched overnight. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-1-the-one-idea-per-sentence-rule]
- Ear version: > We'd been planning the migration since February. It finally launched on Tuesday. The last round of testing turned up two small bugs. We patched them overnight. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-1-the-one-idea-per-sentence-rule]
Core concept 2 — Signposting
- Signpost vocabulary: [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-2-signposting]
- First / second / third / finally - explicit numbering [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-2-signposting]
- Here's the thing - flag a key insight [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-2-signposting]
- Two reasons - set up a numbered list [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-2-signposting]
- Coming up - preview before an ad break [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-2-signposting]
- Back to the story - return marker after a digression [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-2-signposting]
Core concept 3 — No parentheticals, no em-dashed asides
- Before: "The team (which had only formed in January) shipped on time." [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-3-no-parentheticals-no-em-dashed-asides]
- After: "The team shipped on time. They'd only formed in January." [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-3-no-parentheticals-no-em-dashed-asides]
Core concept 5 — The cold open
- Three patterns that work: [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-5-the-cold-open]
- Pattern A - Drop into a scene: > It's 2 a.m. The paging system goes off. Marcus rolls over, reads the alert, and the alert is wrong. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-5-the-cold-open]
- Pattern B - A question with stakes: > What would you do if your entire backup tier disappeared in the middle of a restore? [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-5-the-cold-open]
- Pattern C - A single startling fact: > Last year, 47% of incident retrospectives never produced a single action item. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-5-the-cold-open]
- Anti-pattern - The throat-clear: > Hi everyone, welcome to the podcast. Today we're going to be talking about... [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-5-the-cold-open]
Core concept 7 — Voice-UI prompts (Alexa, Google Assistant, Siri)
- Three constraints beyond "writing for the ear": [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-7-voice-ui-prompts-alexa-google-assistant-siri]
- The prompt must end with an explicit prompt-for-input. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-7-voice-ui-prompts-alexa-google-assistant-siri]
- Confirm without echoing. "Adding milk to your list" is good; "I heard you say milk; I will now add milk to your list" is voice-UI throat-clearing. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-7-voice-ui-prompts-alexa-google-assistant-siri]
- Three-option ceiling. Lists of more than three options exceed working memory. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-7-voice-ui-prompts-alexa-google-assistant-siri]
Core concept 8 — IVR scripts
- Most-likely path first. Order by call-volume share, not alphabetical. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-8-ivr-scripts]
- Always offer a human escape. "Press 0 at any time to reach an agent." [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-8-ivr-scripts]
- Confirm critical inputs. Read back at the cadence a person can write down. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-8-ivr-scripts]
- Re-prompt twice, then escalate. Caller silence twice in a row = route to agent. [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-8-ivr-scripts]
Core concept 10 — Podcast show notes craft
- Required sections: [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-10-podcast-show-notes-craft]
- Episode title - keep under 60 characters; concrete, specific [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-10-podcast-show-notes-craft]
- One-paragraph summary - 50–80 words; uses the episode keyword once [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-10-podcast-show-notes-craft]
- Chapter timestamps - 4–8 chapters; [12:15] How to find your first sponsor [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-10-podcast-show-notes-craft]
- Guest bio - 1 paragraph, links to guest sites [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-10-podcast-show-notes-craft]
- Mentioned resources - every link spoken aloud [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-10-podcast-show-notes-craft]
- Transcript - full or summary [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#core-concept-10-podcast-show-notes-craft]
References
- NPR Training: "How to write a mean script" [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#references]
- Google: Conversation Design [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#references]
- Amazon: Alexa Design Guide [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#references]
- Buzzsprout: "How to Write Podcast Show Notes" [archived source: https://llms-explorer.com/sources/mdb-context-hub/audio-script-writing/#references]
Where this helps
- Writing a podcast script, video voiceover, or narration where the listener gets exactly one pass at the sentence and can't re-scan it the way a reader can. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Designing a voice-UI prompt for Alexa, Google Assistant, or Siri where the response has to end with an explicit prompt-for-input and stay within a three-option working-memory ceiling. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Writing an IVR phone-tree script where call routing needs to be ordered by actual call-volume share and always offer a human escape. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Drafting podcast show notes that need to hit specific structural requirements — a sub-60-character title, a 50-80 word summary, and 4-8 timestamped chapters. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
How to apply this
- Rewrite eye-prose into ear-prose by breaking subordinate clauses into separate short sentences — the pack's own before/after example turns one 40-word sentence with a nested clause into four short, single-idea sentences. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Add explicit signposting vocabulary (first/second/third, 'here's the thing,' 'two reasons,' 'coming up') at exactly the points a reader would otherwise rely on visual structure like headers or bullet points to track. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Cut every parenthetical and em-dashed aside and either drop the information or turn it into its own short sentence, since the ear has no equivalent of parentheses to signal a subordinate thought. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Open with one of the three cold-open patterns that work (drop into a scene, ask a question with stakes, or state a single startling fact) instead of a throat-clearing 'welcome to the podcast, today we're talking about' preamble. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
Antipatterns
- Writing a script with stacked subordinate clauses or parenthetical asides that read fine on the page but lose a listener who gets exactly one pass and can't re-scan. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Opening an episode with a throat-clearing preamble ('Hi everyone, welcome to the podcast...') instead of a cold open that drops into a scene, poses a question with stakes, or states a startling fact. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Writing a voice-UI confirmation that echoes back what it heard instead of confirming without echoing — 'Adding milk to your list' is correct; 'I heard you say milk; I will now add milk to your list' is voice-UI throat-clearing. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Ordering IVR menu options alphabetically instead of by actual call-volume share, forcing the highest-volume caller intent through more menu levels than necessary. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
Known issues
- Listening comprehension speed varies by format — roughly 150-160 wpm for narration but only 110-125 wpm for IVR — so a script tuned for one format's pacing can genuinely be too fast for another. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- A list longer than three items without explicit numbering is a documented comprehension-loss point for listeners, which caps how much a single spoken sentence can safely enumerate. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- IVR scripts need a defined escalation rule (re-prompt twice, then route to a human) — a script with no such rule risks trapping a caller in a loop when their input isn't recognized. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- This pack's reference list is short (four sources: NPR Training, Google Conversation Design, Amazon's Alexa Design Guide, and a Buzzsprout show-notes guide), so deeper platform-specific voice-UI or IVR guidance may need to come from the vendor's own current docs. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
Children
- Cold-Open Pattern (frontier)
- One Idea Per Sentence Rule (frontier)
- Pacing Markers (commas, ellipses, line breaks) (frontier)
- IVR Pause Budgets (100/250/500ms) (frontier)
- Transcription-Friendly Writing (frontier)
- Podcast Show Notes with Chapter Timestamps (frontier)
- Ad-Break Placement (frontier)
- Sample-Dialog for Voice Conversation Design (frontier)
- Broadcast Copy Conventions (frontier)
Frontier under this node: Ad-Break Placement, Broadcast Copy Conventions, Cold-Open Pattern, IVR Pause Budgets (100/250/500ms), One Idea Per Sentence Rule, Pacing Markers (commas, ellipses, line breaks), Podcast Show Notes with Chapter Timestamps, Sample-Dialog for Voice Conversation Design, Transcription-Friendly Writing