How to Design Better NotebookLM Slides (From Default to Brand-Perfect)

NotebookLM's slide generator is one of the most underrated features in the tool. You drop in your sources, hit generate, and out comes a full presentation. Fast. But the default output looks like every other AI-generated deck — plain, generic, and completely disconnected from your brand.

The good news: you can fix that without touching any design software. In this article I'll show you three ways to make your NotebookLM slides actually look good — from the built-in defaults all the way to matching your exact brand colors and typography.

This is also the companion article to my YouTube tutorial, where I walk through each method visually.

---

Method 1: The Default — What NotebookLM Gives You Out of the Box

Before improving anything, it's worth understanding what you're starting with. When you generate slides in NotebookLM, the tool creates a structured Google Slides presentation based on your notebook content. The layout is clean and the structure is solid — it reads your sources and picks the right hierarchy.

The limitation is purely visual. The default theme is a safe, corporate style: light background, dark text, a generic accent color. It works. It's just not yours.

How to generate slides in NotebookLM:

1. Open your notebook and make sure your sources are added
2. Click the Studio panel on the right
3. Select Slides from the output options
4. Hit Generate — NotebookLM will create a Google Slides deck linked to your notebook

That's it. You now have a working deck. For a quick internal presentation or a rough first draft, this is genuinely useful as-is.

When to stop here: If you're sharing internally, presenting to a team that cares about content over design, or using this as a first draft to hand off to a designer — the defaults are fine.

When to keep reading: If the slides represent your brand, are going to clients or an audience, or need to match a visual system you've built — you need one of the next two methods.

---

Method 2: Ask Claude Code to Read Your Codebase and Generate a Design YAML

This is the method that takes about 5 minutes and produces the best results if you already have a website or product with a defined visual identity.

The idea: your brand colors, fonts, and design language already live somewhere — probably in a CSS file, a Tailwind config, or a component library. Instead of manually translating that into a design brief, you ask Claude Code to read those files and output a structured YAML description. You can then feed that YAML into any AI slide tool.

Step 1 — Open Claude Code and use this prompt:

You are a design systems expert. Read my codebase and extract my design system, then output a YAML I can use to prompt AI tools for on-brand slide presentations.

Analyze:
- tailwind.config.ts or globals.css
- Key components: Header, Hero, PricingSection, Footer

Output in this format:

description: "A [mood/style] aesthetic for [use case]. Features [key visual elements]. Powered by [typeface]."
design_system:
  global_style:
    theme: "[Short label]. [One sentence visual language.]"
    typography:
      primary_heading: "[Font name, weight]"
      secondary_heading: "[Font name, weight]"
      body_text: "[Font name, weight]"
    color_palette:
      primary: "#XXXXXX"    # [name]
      background: "#XXXXXX" # [name]
      surface: "#XXXXXX"    # [name]
      text_main: "#XXXXXX"
      text_secondary: "#XXXXXX"
    key_visual_elements:
      - "[Recurring visual pattern]"
      - "[Border or glow treatment]"
      - "[Badge or button style]"
  image_generation_prompts:
    style_guidelines: "[Photography/illustration style matching the design feel]"
    themes:
      - target: "[Slide type — e.g. Hero, Stats, CTA]"
        prompt_elements: "[What to show and how]"
  slide_layout_templates:
    - type: "[Layout_Name]"
      usage: "[Describe: what goes where, what visual elements appear]"

Here are my files: [paste code or attach them]

If you're using Claude Code, you don't need to paste files manually — Claude Code can read your codebase directly. Just open it in your project and ask it to analyze the relevant files.

Example of good output:

Example of good output:

description: "A premium dark SaaS aesthetic with electric violet accents. Features near-black backgrounds, subtle warm radial glows, and gradient-border cards. Powered by Bricolage Grotesque."
design_system:
  global_style:
    theme: "Dark SaaS. Near-black background with violet brand accent and warm glow effects."
    typography:
      primary_heading: "Bricolage Grotesque, black (900), tracking-tight"
      body_text: "Bricolage Grotesque, regular"
    color_palette:
      primary: "#8A5CF5"    # Electric Violet
      background: "#181818" # Near-Black Charcoal
      surface: "#262626"    # Card Surface
      text_main: "#F2F2F2"
      text_secondary: "#999999"
    key_visual_elements:
      - "Thin #363636 border lines framing content columns"
      - "Pill-shaped badges with gradient borders and dark fill"
      - "Ghost lime radial glow in hero corner backgrounds"
  slide_layout_templates:
    - type: "Dark_Hero_Split"
      usage: "Large headline left, product mockup right on near-black with corner glow"
    - type: "Pricing_Tier_Cards"
      usage: "3-column cards with gradient borders — neutral, violet, and soft purple glow"
    - type: "CTA_Dark_Closer"
      usage: "Centered H2, warm-glow CTA button, monospace subcopy below" 

Step 2 — Apply the YAML to generate your slides:

Once you have the YAML, paste it into NotebookLM with this follow-up prompt:

Using the design system in this YAML, generate a slide deck outline for the content below. Apply the color palette, typography, and visual style. For each slide specify: layout type, heading, body copy, and visual element notes.

Design YAML:
[paste your YAML here]

Content to turn into slides:
[paste your notes, outline, or the NotebookLM-generated text here]

This approach works especially well when you already use Kaptex to save your research into NotebookLM — you capture the raw content, NotebookLM structures it, and then you apply your brand design layer on top.

---

Method 3: Screenshot a Design You Like → Drop It Into Gemini → Adapt It

This is the fastest path if you don't have a codebase to reference, or if you want to match a specific slide aesthetic you've seen somewhere.

The idea: take a screenshot of any slide, website, or design you admire, upload it to Gemini (which has native vision capabilities), and ask it to reverse-engineer the design system into a YAML you can reuse.

Step 1 — Capture the design:

Take a screenshot of:

• A slide deck you've seen in a presentation
• A competitor's website
• A product landing page with a visual style you want to borrow
• Any image with a color palette and typography you like

Step 2 — Upload to Gemini and use this prompt:

I'm uploading a screenshot of a design I want to use as inspiration for a slide presentation.

Analyze the visual design and extract:
- Color palette (primary, background, accent, text — hex if possible)
- Typography style (font weight, size relationships, letter spacing feel)
- Key visual elements (borders, shadows, overlays, icons, decorative elements)
- Overall design mood and aesthetic in plain language

Output a YAML design system I can reuse to prompt AI slide tools:

description: "A [mood/style] aesthetic. Features [key elements]. Powered by [typeface]."
design_system:
  global_style:
    theme: "[Short label]. [One sentence visual description.]"
    typography:
      primary_heading: "[Font style, weight]"
      body_text: "[Font style, weight]"
    color_palette:
      primary: "#XXXXXX"
      background: "#XXXXXX"
      surface: "#XXXXXX"
      text_main: "#XXXXXX"
    key_visual_elements:
      - "[Recurring visual treatment]"
      - "[Border, glow, or overlay style]"
  image_generation_prompts:
    style_guidelines: "[Photography/illustration style matching the design]"
    themes:
      - target: "[Slide type]"
        prompt_elements: "[Visual description for AI image generation]"
  slide_layout_templates:
    - type: "[Layout_Name]"
      usage: "[What goes where on this slide]" 

Example of good output:

Example of good output (Tech Pitch style):

description: "A high-energy futuristic pitch deck. Matte black background, neon lime green accents, and grayscale photography. Powered by Open Sauce."
design_system:
  global_style:
    theme: "Tech Pitch. High-contrast startup design with neon green tech accents."
    typography:
      primary_heading: "Open Sauce, bold"
      body_text: "Open Sauce, regular"
    color_palette:
      primary: "#C0FF00"    # Neon Lime
      background: "#000000" # Pure Black
      surface: "#1E1E1E"    # Dark Charcoal
      text_main: "#FFFFFF"
    key_visual_elements:
      - "Neon green diagonal arrows and geometric progress indicators"
      - "Grayscale professional photography with high contrast"
      - "Massive punchy titles that dominate the layout"
  slide_layout_templates:
    - type: "Neon_Pitch_Title"
      usage: "Large title in neon green on pure black with grayscale architectural photo"
    - type: "Impact_Stats_Bold"
      usage: "Huge white numbers with neon green unit labels and grayscale icons"
    - type: "Call_To_Action_Cyber"
      usage: "Massive neon green THANK YOU text with small white contact details in a grid" 

Step 3 — Feed the output back into your slides workflow:

Take the YAML Gemini generates and use the same follow-up prompt from Method 2. The result: a slide deck that matches the visual style of whatever you screenshotted — applied to your own content.

This is particularly powerful for competitive research. If a competitor has a pitch deck aesthetic you want to match, or you saw a conference presentation that felt right for your next investor update — you can reverse-engineer it in under 10 minutes.

---

Bonus: The Kaptex Connection

If you're already using NotebookLM for research, you know the friction: you find something worth saving, you switch tabs, you open NotebookLM, you paste the URL. Over and over.

Kaptex is a NotebookLM Chrome extension that puts a save button directly on every page you visit — YouTube videos, Reddit threads, news articles, research papers — so you can add sources to your notebooks without leaving the tab you're on.

The better your sources, the better your NotebookLM slides. And the faster you can build a notebook, the faster you can get to the design step.

---

Summary

The three methods, ranked by effort:

The YAML format is the key insight: it's a structured, reusable design brief that any AI tool can parse. Once you have one for your brand, you can reuse it every time you generate a deck — no redesigning from scratch, no tweaking colors after the fact.