A 100% browser-based tool to remove watermarks from AI-generated images. Dedicated removers for Google Gemini, ByteDance Doubao (豆包), and Jimeng (即梦), plus a manual brush for Midjourney, Stable Diffusion, signatures or anything else. Calibrated per generator, not a generic eraser. Everything runs locally via MI-GAN ONNX + WebAssembly — your images never leave the tab. No signup, no paywall, no usage limit, no output watermark. 5 languages. MIT licensed. Free forever.
No reviews yetBe the first to leave a review for AI Watermark Remover
Framer AI AgentsDesign and publish professional sites with AI
Promoted
Maker
📌
Hey Product Hunt 👋
I'm shiqi, solo dev behind aiwatermarkremover.dev.
I built this because every "online watermark remover" I tried did at
least one of these:
• Uploaded my image to a server I don't control
• Hid the result behind a paywall after a low-res preview
• Stamped their OWN watermark on the output
• Used a generic inpainter that left obvious smears on known
watermark shapes (Gemini's corner sparkle, Doubao's text strip)
So I wrote dedicated removers for the three AI generators I use most
— Gemini, Doubao (豆包), Jimeng (即梦) — plus a manual brush mode for
Midjourney, SD, signatures, station IDs, anything else.
Everything runs in the browser. The MI-GAN model (~27 MB, Apache-2.0)
loads once, caches, and runs locally via WebAssembly. There is no
backend that could receive your image. On mobile I fall back to a
pure-JS PatchMatch fill so phones don't OOM.
Stack: Next.js 16 static export, ONNX Runtime Web, Cloudflare Pages
+ R2. Source is MIT, ships in 5 languages from day one.
Things I'd genuinely like input on:
• Which generator should I add next? Flux? Recraft? SDXL?
• Anyone seeing watermark styles my detector misses? Drop a sample
in the comments and I'll handle it.
Thanks for taking a look 🙏