If you've ever spent more time writing what to say than building the slides themselves, this is for you. Claude Code can take a bare-bones slide outline and generate a full presenter script — with audience-appropriate tone — in under 30 seconds.
The Problem: You Have Slides, But Nothing to Say
Most presentation prep works backwards. You spend 30 minutes designing a slide, then another hour figuring out what to actually say on it. The content exists in your head — the outline, the bullet points, the structure — but converting that into spoken language is its own painful second job.
The mismatch is a workflow problem, not a creativity problem. You already did the hard thinking when you made the outline. What you need is a way to project that structure into full sentences, fast.
The first thing most people try is writing the script slide by slide, manually. That works, but it scales terribly. For a 15-slide deck, you're looking at 40-60 minutes minimum, and the tone drifts inconsistently between slides.
The Fix: Feed Claude Your Outline File
The core workflow is two steps: drop your slide titles into a plain text file, then pass it to Claude Code with context about your audience.
Step 1 — Create outline.txt
Slide 1: Why we're migrating to microservices
Slide 2: Current monolith pain points
Slide 3: Architecture comparison (before/after)
Slide 4: Migration timeline — 3 phases
Slide 5: Risk mitigation plan
Slide 6: Cost impact and ROI projection
Slide 7: Q&A and next steps
No fancy formatting needed. One slide per line is enough.
Step 2 — Run the command
claude "Read outline.txt and write a 3-5 sentence presenter script for each slide in English. The audience is non-technical C-suite executives."
The audience context is the part most people skip — and it's the part that matters most. "Non-technical executives" produces entirely different vocabulary density than "engineering team leads." Claude adjusts terminology, removes jargon, and shifts the framing toward business outcomes automatically.
What worked for me: I added one extra line specifying the presentation goal.
claude "Read outline.txt and write a 3-5 sentence presenter script for each slide in English. The audience is non-technical C-suite executives. The goal is to get budget approval for a 6-month infrastructure project."
That single goal line made the generated scripts significantly more persuasive — every slide tied back to ROI and risk reduction without me editing anything.
Verification — what good output looks like:
Slide 1 — Why we're migrating to microservices:
Our current system handles roughly 2 million requests per day,
and we're consistently seeing performance degradation during
peak hours. The root cause isn't our team — it's the architecture
itself. Today I want to show you why a targeted migration solves
this, and what it means for uptime and customer satisfaction
over the next two quarters.
Slide 2 — Current monolith pain points:
...
On my Mac, a 10-slide outline produced a complete first draft in about 18 seconds. Manual equivalent: 40+ minutes.
Tone Refinement: Same Content, Different Register
The first draft is rarely the final draft. Claude handles tone adjustments as a follow-up command — no need to regenerate from scratch.
More conversational, with audience engagement:
claude "Rewrite the script you just generated in a more conversational tone. Add at least one direct question to the audience per slide."
Formal and concise for a board meeting:
claude "Rewrite the script in a formal register. Each slide script should be no more than 3 sentences. Eliminate filler phrases."
Simplified for a mixed international audience:
claude "Rewrite the script using plain English. Avoid idioms and acronyms. Aim for a 6th-grade reading level."
The gotcha here: if you chain too many rewrites in one session, Claude can lose track of the original content and start hallucinating slide details. My fix — save each major revision to a numbered file (script_v1.txt, script_v2.txt) before requesting the next pass.
Variations and Gotchas
Slide count sweet spot: Claude handles 8-15 slides cleanly in one pass. Beyond 20 slides, the output starts truncating near the end. Break large decks into two outline files and run them separately.
Bullet points vs. titles only: If your outline has sub-bullets, include them. More input context = more accurate scripts. But even title-only outlines produce usable first drafts.
Language switching: If you're presenting in a language other than English, just specify it in the command. The audience-context framing still applies — just write that in English and Claude figures it out.
claude "Read outline.txt and write the presenter script in Japanese. Audience: engineering managers. Tone: direct and technical."
Environment differences: This workflow runs identically on Mac, Linux, and inside Docker containers with Claude Code installed. No environment-specific config needed — it's just file I/O and a CLI call.
The cue sheet extension: Once the script is done, you can go one step further.
claude "Convert the script into a one-page cue sheet. Format: slide number | key talking point (1 sentence) | transition cue. Bold the most important phrase per slide."
That outputs a printable A4 cue sheet you can hold during the actual presentation — slide transitions marked, key emphasis points bolded, nothing else.
| Task | Manual Time | Claude Code Time |
|---|---|---|
| 10-slide script draft | ~45 min | ~18 sec |
| Tone rewrite | ~20 min | ~10 sec |
| Cue sheet formatting | ~15 min | ~8 sec |
| Total | ~80 min | ~1 min |
Closing
The real value of Claude Code isn't code automation — it's automating repetitive knowledge work. A slide outline is the minimum viable input. The full presenter script, tone variations, and cue sheet come out the other side.
Start with one real outline file you already have. Run the command. See what 18 seconds actually produces before you spend another hour writing scripts by hand.
Next up: using Claude Code to generate post-presentation Q&A prep — anticipated questions and suggested answers — from the same outline file.
---HERO_LABEL_EN: Outline to Script Engine
🐦 Faster updates on X: @baegseungh7061
📚 More in this series: Code Intro
💌 Subscribe: Follow on X or grab the RSS
댓글
댓글 쓰기