| title: SD3.5 Watermark Remover | |
| emoji: 🎨 | |
| colorFrom: purple | |
| colorTo: blue | |
| sdk: gradio | |
| sdk_version: "4.44.1" | |
| app_file: app.py | |
| pinned: false | |
| license: mit | |
| hardware: zero-a100 | |
| # 🎨 SD3.5 Watermark Remover | |
| Remove watermarks from images using **Stable Diffusion 3.5 Large** regeneration, powered by ZeroGPU. | |
| ## How It Works | |
| This tool uses SD3.5's img2img pipeline to regenerate images while preserving their content. By applying controlled denoising, watermarks are effectively removed without manual editing. | |
| ### Key Parameters | |
| - **Strength (0.1-0.6)**: Controls how much the image changes | |
| - Lower values (0.2-0.3): Best for subtle watermarks, preserves most detail | |
| - Higher values (0.4-0.5): Better for prominent watermarks | |
| - **Inference Steps**: More steps = higher quality, slower processing | |
| ## Based On | |
| This Space is based on the [WatermarkAttacker](https://github.com/XuandongZhao/WatermarkAttacker) research project. | |
| ## Model | |
| Uses [Stable Diffusion 3.5 Large](https://huggingface.co/stabilityai/stable-diffusion-3.5-large) from Stability AI. | |
| ## License | |
| MIT License - See the original repository for more details. | |