wmr (gemini-watermark-and-synthid-remover)

Properties

category: watermark-remover
maturity: emerging
approach: Single C++20 binary that auto-detects and removes Google Gemini/Veo visible watermarks via reverse alpha-blending (mathematically inverting the overlay compositing formula rather than inpainting), handles NotebookLM watermarks via AI inpainting (MI-GAN/ONNX), and attempts SynthID (invisible) removal via frequency-domain carrier-signal subtraction or, as a lossy fallback, low-strength SDXL img2img regeneration. Processes both images and video, per-frame.
access: open-source
owns: A cross-platform CLI tool for removing visible and invisible AI watermarks specifically from Google Gemini, Veo, and NotebookLM outputs, including video.
distinctFrom: Unlike remove-ai-watermarks and noai-watermark (both centered on diffusion-regeneration to disrupt pixel watermarks across many vendors), wmr's headline technique for visible marks is a lossless mathematical alpha-blend inversion rather than any generative pass, and it explicitly supports video (Veo) removal rather than images alone.
source: wmr (gemini-watermark-and-synthid-remover)
notes: MIT license, 31 stars; small, single-maintainer, active project (183 commits). Author acknowledges the SynthID (invisible) removal path is less reliable than the visible-watermark path. Exact launch date not confirmed; not previously tracked in this catalogue, flagged as an untracked gap-fill rather than a confirmed brand-new launch within the 2026-07-27 window.

category: watermark-remover
maturity: emerging
approach: Single C++20 binary that auto-detects and removes Google Gemini/Veo visible watermarks via reverse alpha-blending (mathematically inverting the overlay compositing formula rather than inpainting), handles NotebookLM watermarks via AI inpainting (MI-GAN/ONNX), and attempts SynthID (invisible) removal via frequency-domain carrier-signal subtraction or, as a lossy fallback, low-strength SDXL img2img regeneration. Processes both images and video, per-frame.
access: open-source
owns: A cross-platform CLI tool for removing visible and invisible AI watermarks specifically from Google Gemini, Veo, and NotebookLM outputs, including video.
distinctFrom: Unlike remove-ai-watermarks and noai-watermark (both centered on diffusion-regeneration to disrupt pixel watermarks across many vendors), wmr's headline technique for visible marks is a lossless mathematical alpha-blend inversion rather than any generative pass, and it explicitly supports video (Veo) removal rather than images alone.
source: wmr (gemini-watermark-and-synthid-remover)
notes: MIT license, 31 stars; small, single-maintainer, active project (183 commits). Author acknowledges the SynthID (invisible) removal path is less reliable than the visible-watermark path. Exact launch date not confirmed; not previously tracked in this catalogue, flagged as an untracked gap-fill rather than a confirmed brand-new launch within the 2026-07-27 window.