cornelius goodall

Clean Image

Photo-editing

Privacy concerns about cloud-based photo editing have driven innovation in browser-based alternatives, and Clean Image addresses this directly by processing images entirely on the user's device. The tool targets anyone who needs to remove unwanted elements from photos—text overlays, logos, timestamps, blemishes, and stray objects—without uploading their files to external servers. The standout feature is the privacy-first architecture. All processing runs locally through ONNX Runtime and WebAssembly, meaning images never leave the browser. The application uses the OpenCV LaMa neural network to reconstruct areas where users paint over unwanted content, relying on surrounding pixels to generate replacements. This approach works particularly well on clean backgrounds and straightforward removal tasks like text or logos. The user experience emphasizes simplicity. The interface follows an intuitive brush-and-paint metaphor: users tap to aim and paint over elements they want removed, then trigger the erasure. Undo and clear options let them refine before downloading the result. Performance addresses one of the common friction points with local AI tools. The neural network model requires a one-time download of approximately 89 megabytes, which caches in the browser. After this initial load, subsequent uses deliver instant processing with zero data usage, removing the bandwidth burden of repeated cloud uploads. Clean Image supports common image formats including PNG, JPG, and WebP, accommodating most typical photo libraries. The before-and-after comparison feature lets users validate results before saving. The primary limitation stems from its design strength: effectiveness depends heavily on background consistency. Complex, busy backgrounds present challenges for the reconstruction algorithm, which is typical of inpainting-based approaches. Users seeking to remove unwanted content from highly textured or diverse backgrounds would encounter diminishing returns. Clean Image fills a genuine gap for privacy-conscious users who previously had to choose between convenience and data protection. For straightforward removal tasks on suitable images, the local processing approach delivers practical value without sacrificing privacy. The tool serves professionals and casual users alike who prioritize keeping their photos local while leveraging modern AI capabilities.