4 Advanced NotebookLM Slide Deck Tips for Smarter AI Presentations

Ken Fong January 07, 2026

If you’ve scrolled through social media lately, you’ve probably seen NotebookLM’s “one-click deck” demos everywhere. Most people still use it in the most basic way: dump content in, click generate, and accept whatever slide deck comes out.

That’s impressive for a first try, but it’s also exactly why many decks feel generic or “not quite right”.

In this article, we’ll walk through four advanced techniques that give you much more control over NotebookLM’s slide output—so the result better fits your structure, style, and production workflow, instead of the other way around.


1. Use a Slide Outline Document to Control the Content

The biggest pain point with AI-generated decks is:

“I waited for it to finish… and the structure is not what I want at all.”

To fix this, you can separate “planning the slides” from “rendering the slides”.

Step-by-step: Turn an outline into a precise slide deck

  1. Ask NotebookLM to generate an outline from your sources
    Start from your raw materials (articles, notes, reports) and prompt NotebookLM to create a slide outline with:

    • Sections
    • Slide titles
    • Key bullet points per slide

  2. Make the outline a dedicated source
    Export or copy that outline into a new document, and set it as a single reference source inside your Notebook.
    Rename it clearly as 《投影片大綱》 (Slide Outline) so you can spot it instantly.

  3. Switch to Presentation Mode and lock the behavior with a strict prompt
    In presentation editing mode, use a clear instruction such as:

    「請嚴格按照《投影片大綱》文件中的指示，將整份大綱轉化為視覺化的投影片。」
    (“Strictly follow the instructions in the ‘Slide Outline’ document and turn the entire outline into visual slides.”)

This way, you treat the outline as the “single source of truth”, and NotebookLM acts more like a designer than a strategist.

Now NotebookLM is no longer guessing your structure. It’s forced to respect your outline, slide by slide—so the only thing it needs to decide is visual composition, not content hierarchy.

2. Lock in a Visual Style Using a Background or Brand Reference

Even when the content structure is right, the visual style of AI slides can still feel off-brand: colors are weird, fonts don’t match your usual look, backgrounds feel random.

To fix that, treat design style as a data source, just like text.

Step-by-step: Use a background image as a style guide

  1. Find or create a background you love
    Use resources like SlidesGo / Freepik, or generate a custom slide background with your favorite AI image tool.

  2. Upload that background into your Notebook
    Add it as a new file and rename it clearly as 《背景》 (Background).

This is especially useful if you’re creating decks for a brand, course, or product launch where visual consistency really matters.

  1. In Presentation Mode, lock the design prompt
    In the instruction area, include something like:

    「請使用來源中的《背景》圖片作為整份簡報的主視覺風格（不要更改配色、元素類型）。」
    (“Please use the ‘Background’ image from the sources as the main visual style for the entire deck. Do not change its color scheme or visual elements.”)

From this point, NotebookLM will try to align all slides with your reference background—colors, shapes, and overall vibe—so your deck looks coherent instead of “AI-random”.

Pro tip:
Used together, the outline controls what you say and the background controls how it looks—so you spend less time fixing random AI design choices later.

  • the outline controls what to say
  • the background controls how it looks

3. Work Around Slide Count Limits by Splitting the Deck

NotebookLM lets you pick between short / default / long deck lengths in custom mode, but the exact number of slides still depends on:

  • how much content you feed it
  • how it chooses to chunk that content

Almost everyone eventually asks: “Can I just tell NotebookLM to make exactly 50 or 60 slides?”

You can try prompts like “Please create a 65-slide presentation”, but in practice, the slide count is often unstable. In my own tests, simply asking for “65 slides” produced inconsistent results, so I switched to splitting the outline.

If you want more precise control, especially for very long decks, the practical solution is to generate in parts.

Step-by-step: Split large decks into multiple passes

  1. Split your master outline into two parts
    For example:

    • Part 1: Intro + Techniques 1–2
    • Part 2: Techniques 3–4 + Case Studies / Q&A

  2. Generate slides for the first half
    In NotebookLM, select only the first-half outline as the source and generate the deck (this becomes your “Part 1” slides).

  3. Upload the “Part 1” slides as a new source
    Export that deck as PDF, then upload it back to the Notebook as a new reference file. This gives NotebookLM a visual reference of your style and layout decisions.

  4. Generate the second half with visual continuity
    Now, for Part 2:

    • Select both the Part-1 slide deck and the second-half outline as sources.
    • Ask NotebookLM to generate a new deck that continues the style and structure:

    「請參考已生成簡報與下半部分大綱，延續相同版面與風格完成下半部簡報。」

This “two-phase” method:

  • Bypasses length constraints
  • Keeps visual style consistent across both halves
  • Gives you more manual control over where the deck starts and ends

Over time, this “modular outline” approach also makes it easier to reuse sections across different talks and workshops.

4. Edit and Optimize NotebookLM Slide Decks After Export

Here’s the biggest frustration:
There’s a clear workflow gap between “AI-generated slides as images” and “client-ready slides you can fully control.”
NotebookLM currently generates slides as images inside a PDF.

That means:

  • Converting to PDF and dropping into a “PDF editor” doesn’t magically give you editable text.
  • What you’re seeing is still essentially a picture of a slide, not a real PowerPoint textbox.

So how do you tweak:

  • a single typo?
  • a slightly misaligned title?
  • a bullet that needs one more keyword?

Right now, there are two main approaches.

Method 1: Canva – Magic Grab & Grab Text (Pro)

In one line: Use Canva Pro to extract text (Grab Text) and fix layout (Magic Grab) so your NotebookLM PDF becomes an editable deck.

If you upload your NotebookLM PDF into Canva, there are now two Pro features that matter:

  • Magic Grab – separates key elements (including text blocks) from the background so you can move and resize them, mainly for layout control.
  • Grab Text – detects and extracts text inside an image, and converts it into editable text objects on the canvas.

A practical workflow looks like this:

  1. Upload the full NotebookLM-generated PDF to Canva.
  2. Open it in the editor and, for each slide image, use Grab Text to pull the words out of the picture and turn them into real text boxes.
  3. Adjust the wording directly in those text boxes.
  4. Use Magic Grab where needed to move text blocks or other elements for better layout.

Limitations:

  • Fonts may change during extraction, so you’ll usually need to reapply your brand fonts afterward.
  • Complex layouts may require manual clean-up and re-alignment.

So you can think of it this way:

  • Magic Grab = layout flexibility
  • Grab Text = real text editing

Used together, they turn a static NotebookLM PDF into a deck you can meaningfully rework inside Canva.

Method 2: Lovart – Edit Element Workflow

In one line: Use Lovart’s Edit Element to turn static slide images into editable text boxes when you need fine-grained control.

If you need another route to fully edit both text and layout, Lovart’s Edit Element feature gives you deep control.

  1. Convert the NotebookLM PDF to images
    Use any PDF-to-image converter so each slide becomes a standalone PNG/JPG.

  2. Upload these images to Lovart
    Open them inside Lovart and use the 「編輯元素」 (Edit Element) feature.

  3. Turn “text as picture” into editable text boxes
    Lovart will attempt to:

    • detect text regions
    • convert them into editable text elements
    • let you adjust the wording directly

Caveats:

  • The resulting fonts may not perfectly match the original.
  • Sometimes bullet points or icons remain in the background layer, not in the extracted text.

When that happens, you can use Lovart’s 「消除工具」 (Erase/Cleanup tool) to manually remove stray bullet icons or artifacts.

In my own workflow, Canva is my first stop for quick text edits, and Lovart is my choice when I need pixel-level control.

This workflow is more tedious, but it’s currently one of the most practical ways to:

  • start from an AI-generated, image-based slide deck
  • end with fully editable, post-processed slides suitable for client work or professional delivery

5. NotebookLM Slide Deck FAQ 

Q1. What is NotebookLM’s Slide Deck feature and what can it generate?
NotebookLM’s Slide Deck feature converts your notebook sources (notes, articles, outlines, brandbooks) into AI-generated slide decks. You can customize the deck length (short / default / long), choose between presenter-style or detailed slides, and guide the content and design using prompts and reference files.

Q2. Can NotebookLM export slides directly to PowerPoint or Google Slides?
Right now, NotebookLM exports slide decks as PDF files that contain image-based slides. It does not yet export directly to native PowerPoint (PPTX) or Google Slides formats, which is why workflows with Canva or Lovart are helpful for further editing.

Q3. How do I keep NotebookLM slides on brand?
Use a dedicated style reference, such as a brandbook or slide background image, and upload it as a source (for example, a file named 《背景》). Then, in the Slide Deck prompt, tell NotebookLM to follow that visual style and color palette. Combining this with a detailed slide outline gives you on-brand, well-structured decks.

Q4. How can I create long-form NotebookLM presentations that exceed the default page limits?
Split your outline into multiple sections and generate the deck in parts. First, create slides for Part 1 using only the first half of the outline. Export that as a PDF and re-upload it as a visual reference, then generate Part 2 using both the Part-1 deck and the second-half outline as sources. This preserves visual consistency while bypassing slide count limits.

Q5. What’s the best way to edit text in NotebookLM-generated PDFs?
The two most common workflows are:
Canva Pro – use Grab Text to extract text from slide images into editable text boxes, and Magic Grab to adjust layout.
Lovart – convert PDFs to images, then use Edit Element to turn static text regions into editable text boxes and clean up remaining artifacts with the erase tool.

Q6. Who should use NotebookLM’s Slide Deck feature in their workflow?
NotebookLM’s Slide Deck feature is ideal for creators, educators, and solo business owners who already have structured content (articles, lesson plans, playbooks) and want to turn it into consistent slide decks quickly. If you regularly repeat similar workshops or talks, the outline-plus-style-reference method described in this article can save you hours per presentation.

6. Traditional Chinese Summary（約 800 字）

這幾天，大家應該都被 NotebookLM 的「一鍵生成簡報」洗版了。多數人的使用方式很直接：把文章、筆記、報告通通丟進去，請 NotebookLM 幫忙生一份簡報，覺得版面漂亮就收工。但當 AI 生成工具變成人人都會用的東西時，只停留在這個層級，作品在觀眾眼中很快就會失去辨識度與價值。

本文分享四個進階技巧，幫助你把 NotebookLM 當成「專業簡報輔助工具」，而不是只會按一個按鈕的玩具。

一、用《投影片大綱》精準控制簡報內容（避免 AI 亂拆主題）
第一個技巧，是把「規劃簡報」與「生成簡報」拆開來做。你可以先請 NotebookLM 依照原始資料產生一份投影片大綱，接著把這份大綱獨立成一個檔案，並在 Notebook 中重新命名為《投影片大綱》，設為唯一的參考來源。之後在簡報模式下下指令：「請嚴格按照《投影片大綱》文件中的指示，將整份大綱轉化為視覺化的投影片。」這樣 NotebookLM 就不再亂猜內容架構，而是乖乖按照你規劃好的章節與頁數來排版，只負責處理視覺呈現。

二、用《背景》或品牌素材鎖定整體視覺風格（維持品牌一致性）
第二個技巧是掌控簡報風格。許多 AI 產生的簡報看起來「不太像你的作品」，問題通常出在配色、背景與元素風格太隨機。你可以先到 SlidesGo、Freepik 找到喜歡的模板背景，或用 AI 生成一張專屬主視覺，然後把這張圖片上傳到 Notebook，命名為《背景》。接著在簡報模式的指令中寫上：「請使用來源中的《背景》圖片作為整份簡報的主視覺風格（不要更改配色、元素類型）。」NotebookLM 在生成每一頁時，就會主動參考這張背景，在顏色、元素類型與整體風格上保持一致。若再搭配《投影片大綱》，你就同時掌握了「說什麼」與「長什麼樣」。

三、突破 NotebookLM 頁數限制，分批生成長篇簡報（適合課程與工作坊）
NotebookLM 雖然可以選擇簡報長度（短／預設／長），但實際頁數仍受到內容多寡與模型切分方式影響。即使你在提示詞加上「請生成 65 頁簡報」，結果仍然不一定穩定。如果你想做的是長篇課程、講座或工作坊教材，一個實用的做法是「分段生成」。具體做法是先把《投影片大綱》拆成上下兩部分，只勾選上半部，生成第一份簡報；接著把這份上半部簡報輸出為 PDF 後再上傳回 Notebook 當作新來源，與下半部大綱一起提供給 NotebookLM，要求它在延續相同版面與風格的前提下完成下半部。這樣不但可以繞過頁數上限問題，也能確保前後簡報在視覺與結構上連貫一致。

四、解決 NotebookLM「圖片式」簡報的後製問題（讓簡報真的可編輯）
目前 NotebookLM 生成的簡報，本質上仍是「圖片檔」，即使用一般 PDF 編輯工具，也無法直接編輯其中的文字內容。要微調字句、挪動標題位置或修正文案，就必須借助其他工具。第一種方式是使用 Canva Pro 的功能組合：先用 Grab Text 將 NotebookLM 投影片中的文字從圖片裡「抓」出來，轉成真正可編輯的文字框，再搭配 Magic Grab 去移動與調整版面。這樣一來，你就可以在 Canva 裡同時做到「改文字」與「調版面」，最後再輸出成 PPTX 或其他格式。

另一種方式則是使用 Lovart 的「編輯元素」。做法是先把 NotebookLM 的 PDF 轉成單張圖片，再把每一張投影片圖片上傳到 Lovart，使用「編輯元素」將原本當作圖片處理的文字，拆解成可編輯的文字框。雖然字型不一定完全一致，有時候項目符號會被留在背景中，需要再用「消除工具」清除，但整體來說，已經可以讓你在 NotebookLM 的基礎上做更細緻的後製調整。

真正拉開差距的，不再是「會不會用 NotebookLM」，而是「能不能比別人更聰明地使用 NotebookLM」。如果你願意多花一點心思，掌握上述四種進階使用方式，就能讓 AI 幫你省下大量時間的同時，仍保留作品該有的專業質感與個人風格。

下次再打開 NotebookLM，要不要試著套用其中一兩個技巧，看看到底能替你省下多少修改時間？

References

  • NotebookLM official product documentation
  • Google: 8 ways to make the most out of Slide Decks in NotebookLM
  • Canva – Presentation design, Magic Grab & Grab Text features
  • Lovart – Image and element editing features

Keywords

NotebookLM advanced tips, AI presentation workflow, Google NotebookLM slides, AI-generated slide deck editing, Canva Magic Grab, Canva Grab Text, Lovart Edit Element, AI-powered slide design, presentation design automation, long-form AI presentations, slide outline strategy

— Dr. Ken FONG

Tags
Ken Fong

Posted by Ken Fong

Post a Comment

0 Comments