Reply To: Can AI generate speaker notes from slide bullet points?

Yes — your five‑minute, one‑slide test is the right move. It surfaces the real issues fast: tone drift, invented details, and timing. Let me add a simple calibration trick and a reusable template that will tighten your results on the first try.

Quick context

AI is excellent at turning tidy bullets into usable speaker notes. The two levers that make it presentation‑ready are: 1) matching your speaking speed so timing lands, and 2) giving the AI a clear structure to follow so the voice stays consistent.

What you’ll need

• One slide with 3–6 clean bullets (mark gaps as [missing]).
• Audience and tone (e.g., executive, conversational, training).
• Target time per slide (e.g., 90 seconds).
• Your speaking speed (quick test below) and a timer.

Calibrate your timing in 3 minutes

1. Read a 120‑word paragraph aloud at a natural pace. Time it.
2. Words per minute (WPM) = 120 ÷ seconds × 60. Most presenters sit between 130–160 WPM.
3. Target word count = (Target seconds ÷ 60) × WPM. Example: 90s × 140 WPM ≈ 210 words.

Copy‑paste prompt (primary)

“You are my presentation coach. Turn the slide bullets into speaker notes using the Beat‑Map structure. Audience: [describe]. Tone: [conversational/executive/training]. Target time: [seconds]. My speaking speed: [WPM]. Target word count: [number] words; stay within ±5%. Do not invent facts; if something’s missing, write a bracketed placeholder like [insert stat]. Output in plain text, no markdown.

Beat‑Map structure to follow:

1) Headline (max 7 words)

2) Three beats (one short sentence each) with [2s pause] between beats

3) One quick example or analogy tailored to the audience

4) One call‑to‑action or key takeaway (one sentence)

5) One‑line transition to the next slide

Also include at the end:

– A checklist of factual claims to verify (no sources, just list the claims)

– A 2‑line style summary of the voice you used”

Fast variants you can swap in

• Executive: “60 seconds, no examples, numbers first, remove adjectives, keep sentences under 12 words.”
• Training: “120 seconds, include one analogy and two audience questions.”
• Teleprompter layout: “Line breaks every 8–12 words; group into 3 chunks; include [PAUSE] cues; bold 3 keywords per chunk.”
• Data‑sensitive: “No percentages unless provided. If comparison needed, say ‘higher’ or ‘lower’ without numbers.”
• Timing fix: “Shorten by 15% without losing the key takeaway; keep the headline.”

Step‑by‑step workflow (10–20 minutes for 3 slides)

1. Clean the bullets: trim vague words; add [missing] markers where data is needed.
2. Run the primary prompt for one slide using your WPM and target words.
3. Read aloud with a timer. If you’re off by more than 10%, ask the AI to expand or compress by a specific word count.
4. Lock the voice: extract a 2–3 line style brief from the best output (pace, tone, words to avoid). Reuse it for every slide.
5. Bulk‑generate 2–3 more slides. Keep the “verify claims” checklist attached to each.
6. Final pass: add one personal line per slide (an example, a client moment, or a contrast) so it sounds like you, not a script.

What good output looks like

• About the target word count you set (e.g., ~210 words for 90 seconds at 140 WPM).
• Short sentences, clear beats, and a visible [2s pause] cue so you can breathe.
• One concrete example and a crisp transition to the next slide.
• A small checklist of claims you’ll verify before presenting.

Insider trick: the 1–1–1 polish

• One emphasis word: Bold a single word per beat; it anchors attention.
• One personal tag: Insert [MY STORY: …] where you’ll add a real moment.
• One breath cue: Keep every third sentence under eight words.

Example bullets (for practice)

• Customer churn down to 8% [verify].
• Top driver: faster onboarding (from 14 to 7 days) [verify].
• Secondary driver: clearer pricing page [missing comparison].
• Next step: pilot the new flow with 3 accounts.

Run the prompt with those bullets, set 90 seconds, 140 WPM, and ask for teleprompter layout if you present from notes.

Common mistakes and easy fixes

• Tone creep across slides. Fix: reuse a 2–3 line style brief in every prompt.
• Timing misses. Fix: calculate word targets up front; ask the AI to add or remove N words, not “make shorter.”
• Invented specifics. Fix: require “[placeholder]” for missing facts and include the verification checklist.
• Dense paragraphs. Fix: request short sentences and [2s pause] between beats; use teleprompter layout.
• Flat delivery. Fix: add one audience question or a micro‑contrast (“before → after”) per slide.

45‑minute sprint plan

1. 5 min: WPM calibration + word targets for 3–5 slides.
2. 15 min: Generate drafts with the primary prompt (include the style brief).
3. 15 min: Verify claims, add your personal line, tighten timing by ±10%.
4. 10 min: Rehearse once, mark pauses, and smooth transitions.

One extra prompt (quality control)

“From the script above, list every factual claim and what evidence is needed to confirm it. Do not add sources. Output as a simple checklist I can verify.”

Closing thought

AI can absolutely draft strong speaker notes from bullets. Calibrate your timing, give it a clear Beat‑Map, and reuse a tiny style brief. You’ll keep your voice, hit your time, and cut prep by more than half.