NotebookLM Watermark Removal Guide: Free Alternatives to the $250/Month Ultra Plan
NotebookLM Watermark Removal Guide: Free Alternatives to the $250/Month Ultra Plan
One Watermark, One Broken Workflow
Anyone doing research with NotebookLM knows the feeling: the tool is genuinely powerful — feed it a pile of PDFs, web pages, and YouTube links, and it produces polished podcast-style audio overviews, slide decks, video overviews, and infographics. But the free tier has one unavoidable catch: every export carries a watermark. A fixed logo sits in the bottom-right of videos, a brand mark in the corner of PDF and PPTX exports, another on infographics. Want to present those slides, submit that deck, or upload that video to YouTube? First you have to explain why there's an ugly badge on screen.
The official fix is an AI Ultra subscription at roughly $250/month — a price almost nobody will pay just to remove a watermark. Fortunately, by 2026 a whole free ecosystem has grown up around NotebookLM watermark removal: browser-local online tools, open-source CLIs, and dedicated PDF scripts. This guide covers cleaning methods for all 8 export formats, compares the tools, and explains the clever math behind the best techniques.
One note before we start: in July 2026 Google renamed NotebookLM to Gemini Notebook. Existing exports, watermark positions, and cleaning methods are unchanged.
1. Know Your Enemy: Watermark Positions & Sizes
To clean a watermark you first need to know where it lives. Community reverse-engineering shows NotebookLM watermarks appear at highly predictable positions:
| Format | Position | Size (1080p reference) |
|---|---|---|
| Video | Bottom-right | x=1104, y=656, w=770, h=62 |
| Video (720p) | Bottom-right | x=736, y=437, w=513, h=41 |
| PDF/PPTX | Bottom-right corner | ~350×80px scan area |
| Infographic | Bottom-right corner | Variable |
| Video ending | Last 2.5 seconds | "Made with Google" screen |
The pattern is clear: every mark sits in the bottom-right corner, and videos end with a 2.5-second official outro. Fixed positions are a gift for automation — tools can batch-process precisely instead of re-locating the mark on every file.
2. Option 1: NotebookLM Remover — Browser-Local All-Rounder
The community-maintained notebooklmremover.org is currently the most format-complete free tool, covering 8 export types:
- Video watermark (MP4 from NotebookLM Video Overview)
- PDF slide decks
- PPTX presentations (animations and notes preserved)
- Infographics (PNG / JPG / WebP)
- Gemini visible marks (on Gemini-generated images)
- Audio intro/outro trimming (Audio Overview podcasts)
- Image metadata cleanup (EXIF / XMP)
Its biggest selling point is 100% browser-local processing: the whole tool runs in your browser on WebAssembly, files are never uploaded to a server, and you download the result directly. For unreleased research or commercial proposals, that privacy guarantee is decisive.
The workflow is three steps: upload → auto-process (PDF/PPTX/infographics are cleaned in-browser; videos use common export presets plus optional ending trim) → download.
3. Option 2: The Open-Source CLI — notebooklm-remover
For command-line lovers and batch processing, the npm package notebooklm-remover does the job:
npm install -g notebooklm-remover
The technical approach differs from the online tools: it uses connected component analysis with adaptive thresholding to detect and cleanly fill the watermark region. It works on light and dark backgrounds and handles both the NotebookLM text logo and icon variations. Supports PNG/JPG/WebP.
Useful options:
| Flag | Description | Default |
|---|---|---|
--threshold <n> | Detection sensitivity | 60 |
--scan-w <pct> | Scan width ratio from right edge | 0.22 |
--scan-h <pct> | Scan height ratio from bottom edge | 0.08 |
Know its boundary: this CLI handles images only (slide screenshots, infographics). For video, PDF, PPTX, and audio, use the online tool above (maintained by the same team, also fully browser-local). MIT licensed, free to integrate.
4. Option 3: Other Tools Worth Knowing
- SlideClean (open source on GitHub): local Python script specialized in PDF slide watermarks — a good fit for academic users who don't want to upload files
- SlideDeckcleaner.com: online PDF processing, free, server-side (skip if you're privacy-sensitive)
- SlideSweep: claims "seconds, not minutes"; AI auto-detects and masks "Made with NotebookLM" branding; free, no sign-up, files deleted immediately after processing
- DrWatermark (watermark.phd): AI-powered online tool for video and PDF; real research-team usage reports mention fast processing and clean output
- Tenorshare PDNob: takes the OCR route — runs OCR on PDF slides to unlock the text layer, then lets you select and delete the watermark directly; better if you also need to edit slide content
- CapCut manual method: drag the video into CapCut and cover the fixed bottom-right watermark with a blur/mask; free but manual per video and useless for moving watermarks
5. Pro Tip: Lossless Math Removal for Gemini Image Watermarks
The coolest technical detail in the NotebookLM ecosystem: Gemini-generated images carry a visible sparkle watermark that can be removed with mathematically lossless alpha-channel reversal — no AI inpainting, no quality loss:
original_pixel = (watermarked_pixel - α × 255) / (1 - α)
The key insight is that the watermark is semi-transparent and composited with a standard alpha-blend formula. If you know the alpha template, you can exactly recover the original pixels underneath. Template parameters:
| Image Size | Alpha Template |
|---|---|
| ≤1024px | 48px margins |
| >1024px | 96px margins |
| Standard sizes | 1024×1024, 1536×1024, 2816×1536 |
This is also why local tools like OpenNoMark (covered in our previous guide) can do "reverse-alpha recovery" on Gemini images — it isn't guessing, it's exact restoration.
6. Don't Forget Metadata: C2PA / EXIF / XMP
AI-generated images carry provenance in their metadata (C2PA, EXIF, XMP). If removing the visible watermark isn't enough, clean the metadata too — two free one-liners:
# exiftool full wipe
exiftool -all= image.jpg
# or Python PIL re-save
from PIL import Image
img = Image.open("in.jpg")
img.save("out.jpg")
7. Free vs Ultra: Who Should Pay?
| Item | Free tier | AI Ultra (~$250/mo) | NotebookLM Remover |
|---|---|---|---|
| Slide watermark | Yes | No | Removed free |
| Video watermark | Yes | No | Removed free |
| "Made with Google" ending | Yes | No | Trimmed free |
| Privacy | Server-side | Server-side | 100% browser-local |
| Cost | Free | $250/month | Free |
Recommendations:
- Occasional slide/infographic exports → browser-local tool; one minute, zero cost
- High-frequency video overview processing → try local tools first; only evaluate Ultra if volume justifies it
- Unreleased research or trade secrets → stick to 100% local options (browser-local or open-source CLI); files never leave your device
- Need to edit slide content itself → the PDNob OCR route fits better
8. Summary
The NotebookLM (now Gemini Notebook) watermark problem is fundamentally a conflict between an overpriced official solution and a restricted free tier. By 2026 the free tool ecosystem has resolved most of it: a browser-local tool covers all 8 formats with full privacy, an open-source CLI covers batch automation, PDF-specific tools and OCR routes each have their niche, and Gemini images can even be cleaned losslessly with math.
Bottom line: run your exports through a browser-local tool first — odds are you never need to pay a cent for watermark removal. Consider the open-source CLI or an Ultra subscription only if you process at very high volume and need automation.