NotebookLM's PPTX Export Is Images — How to Get Editable Slides

2026-08-26

TL;DR— NotebookLM's PowerPoint export (added February 2026) saves each slide as a single flat image. You cannot select, retype, or restyle the text in PowerPoint. To actually edit the slides, you need a conversion step that separates the text from the background — that is what Swift-Slide does, keeping your wording exactly as generated.

What NotebookLM's PPTX export actually gives you

Since the February 2026 update, NotebookLM (Gemini Notebook) can export its generated slides as a .pptx file, and you can revise slides with prompts before exporting. That is genuinely useful — but when you open the file in Microsoft PowerPoint, each slide is one baked-in image. Clicking the headline selects a picture, not text. Fixing a single typo, changing a number, or matching your corporate font still means rebuilding the slide by hand.

Your options for making the slides editable

1. Rebuild manually. Re-create the layout in PowerPoint and retype everything. Faithful, but slow — infographic-style slides can take hours.

2. Generic OCR. Copy the text out with an OCR tool and paste it into a blank deck. You get the words, but you lose the design entirely.

3. AI conversion that preserves the design. A converter reads the text, removes it from the image, restores the background behind it, and places the same text back as editable text objects — so the slide looks the same but is now editable.

How Swift-Slide converts NotebookLM slides

Swift-Slide takes the third approach. Upload the PDF or image that NotebookLM generated, and AI lifts the text off, inpaints the background, and re-places the identical text as editable objects. Edit in the browser, then export a PPTX that opens natively in PowerPoint.

Before: an infographic slide generated by NotebookLM, where text cannot be edited — Before (image only)

After: the same slide converted by Swift-Slide, with the same text re-placed as editable objects — After (same text, repositioned as editable objects)

Actual conversion results. Output quality varies with the source material.

Conversion takes roughly 30–60 seconds per image, with multi-page files processed in parallel. There is a free plan (5 pages per month, no credit card), so you can test it on your own slides first.

*Times are our measured examples (Mar–Jul 2026). Results vary with source material and load.

FAQ

Does converting change my slide content?

No. Swift-Slide reads the text on your slide, lifts it off, restores the background behind it, and places the same text back as editable text objects. The wording stays identical — it just becomes editable.

Does it work with slides in languages other than English?

Yes. Japanese and English are fully supported, and the OCR engine handles mixed-language slides.

What happens to my uploaded file?

Uploaded PDFs and images are deleted promptly after processing (zero-retention design), and your documents are never used to train AI models.

Try it on a NotebookLM slide — no sign-up needed

Open the demo editor

Or see how it works on the Swift-Slide top page.

Swift-Slide is not affiliated with Google LLC or NotebookLM (Gemini Notebook). NotebookLM is a trademark of Google LLC. Feature descriptions of NotebookLM are accurate to the best of our knowledge as of August 2026.