About NotebookLM Remover
A free, browser-local cleanup tool for AI-generated exports. Every file is processed inside your own browser tab — nothing is uploaded, and there is no server to upload to.
What NotebookLM Remover is
NotebookLM Remover is an independent web tool that removes the visible watermark Google stamps onto files exported from NotebookLM (now also branded as Gemini Notebook) and Gemini. It runs entirely in the browser: you open a page, drop in a file, and get a cleaned copy back without creating an account, installing anything, or waiting in a queue.
There are seven tools on the site, each tuned to a different export format:
- Video remover — clears the badge burned into exported Video Overviews.
- PDF slides remover — cleans slide decks exported as PDF.
- PPTX remover — cleans PowerPoint decks without flattening them into images.
- Infographic remover — handles the tall, single-image infographic exports.
- Gemini image remover — reverses the sparkle mark on Gemini-generated images.
- Podcast trimmer — trims the spoken sign-off from Audio Overviews.
- Metadata cleaner — strips EXIF and other non-pixel data from images.
How it works, technically
Each format needs a different approach, and the details are worth stating plainly — they are the best evidence that this is a real piece of engineering rather than a wrapper around someone else's API.
Video
FFmpeg is compiled to WebAssembly and executed inside the browser tab. The watermark is removed with FFmpeg's delogo filter applied over preset coordinates measured from real exports — for 1080p output that is a 770×62 pixel box at x=1104, y=656, with a separate coordinate set for 720p. There is an optional 2.5-second tail trim that drops the closing "Made with Google" card, plus optional frame-rate and colour-grade passes.
PDF slides
pdf.js renders each page to a canvas at scale 2 (scale 3 in high-quality mode), the watermark is erased on that canvas, and pdf-lib rebuilds a fresh PDF from the cleaned pages. Because the pages are re-rendered, the output is image-based — that is a deliberate trade-off, and it is why the PPTX tool exists separately.
PPTX
A .pptx file is a ZIP archive. JSZip unpacks it, only the images under ppt/media/ are run through the image engine, and the archive is repacked. Slide text, layout, and shapes are never touched, so the deck stays fully editable in PowerPoint or Google Slides.
Images and infographics
The engine scans the bottom-right corner of the image — a region of at least 350×80 pixels that scales up with the image dimensions — and converts it to grayscale. Connected-component analysis then isolates the dark pixel clusters that make up the mark, discarding anything that looks like a border, a thin rule, or ordinary content. The recovered area is filled by interpolating a vertical gradient from the untouched pixels above and below it, which reconstructs the background instead of pasting a flat rectangle over it.
Gemini images
Gemini's sparkle mark is composited at a known opacity, which makes it reversible rather than merely paintable. The engine reads the alpha template for the image size and inverts the blend per channel: original = (watermarked − α × 255) / (1 − α). On the standard Gemini output sizes this is mathematically lossless — it recovers the exact original pixel values, not an approximation. Non-standard sizes fall back to detecting the sparkle centre and inpainting, which is a genuine estimate.
Audio
The podcast trimmer cuts the tail using an FFmpeg stream copy (-c copy). The audio is never decoded and re-encoded, so trimming costs zero generation loss — the samples you keep are bit-for-bit identical to the ones in the source file.
Metadata
The metadata cleaner draws the image onto a canvas and re-encodes it. This removes EXIF, GPS, and every other non-pixel segment by construction, because a canvas only carries pixels — there is no metadata left for the encoder to write back out.
Why nothing is uploaded
This is an architectural fact, not a privacy promise. The site is a static bundle of HTML, CSS, and JavaScript. There is no upload endpoint, no file storage, no account system, and no server-side processing to send anything to. Even if we wanted your files, the application has nowhere to put them.
You do not have to take that on faith. Open your browser's developer tools, switch to the Network panel, and clean a file: you will see the page's own assets load, and then nothing. No request carries your file anywhere. We would rather you check than believe us.
Why it is free
Because it costs almost nothing to run. Server-side watermark removal means paying for CPU and bandwidth on every single file, which forces credit limits, sign-ups, and paid tiers. Here the work happens on your own machine, so serving one user and serving a hundred thousand cost the site roughly the same: static file delivery. There is no per-file bill to pass on to you.
The site carries Google AdSense advertising to cover the domain, hosting, and development time. That is the whole funding model — there is no paid plan, no upsell, no watermarked free tier, and no data to sell, because your files never arrive here in the first place.
What this tool deliberately does not do
It removes visible marks only. It does not remove SynthID or any other invisible or cryptographic watermark — those are embedded in the pixel and audio data itself, survive the processing described above, and remain detectable afterwards. It does not break DRM, bypass paywalls, or unlock paid features.
It also does not change anything about ownership. Removing a badge from a file changes how the file looks, not who holds rights to the underlying content or how you are permitted to use it. Please use this on exports you generated yourself or are otherwise authorised to process. Our acceptable use policy spells out where that line sits.
Who it is for
Students turning an AI-generated study deck into something they can hand in. Teachers building lesson material. Researchers folding an Audio Overview into a literature review. Presenters who need a clean slide for a conference screen. Creators and knowledge workers who use NotebookLM and Gemini as drafting tools and want the finished artefact to look like their own work rather than a tool's output. In every one of those cases the file already belongs to the person cleaning it.
An independent project
NotebookLM Remover is not affiliated with, endorsed by, sponsored by, or connected to Google, NotebookLM, Gemini Notebook, or Gemini in any way. "NotebookLM", "Gemini", and "Google" are trademarks of Google LLC and are used here only to describe the file formats this tool works with. We have no partnership, no API access, and no inside knowledge — the engines described above were built by examining real exported files.
Contact and corrections
If something on this site is wrong, out of date, or does not work the way this page describes, we want to hear about it. Reach us through the contact page. Security issues — anything about how the client-side code behaves or what it touches — belong on the security page, which explains how to report them. You can read about who maintains the project and how articles here are researched on the maintainer page.