v1.1.1 - Some Stability Fixes

[1.1.1] - Some Stability Fixes

🐛 Fixed

• Fixed an issue where Upscaler could hang indefinitely during model download when the network stalled.
• Fixed an issue where Upscaler could remain stuck during ONNX session creation or inference without timing out.
• Added timeout and abort handling for model download, model stream reads, session initialization, and inference execution.

🔄 Changed

• Changed the default remover model to ormbg_quantized.
• Updated BiRefNet Lite model URLs to use the new BiRefNet_lite-ONNX repository from ONNX Community.
• Renamed birefnet_lite_quantized to birefnet_lite (FP32) with updated size (224 MB) and description for standard precision without fidelity loss.

🧠 Improved

• Add border to image settings panel for better component organization
• Optimized image compression by switching from PNG to WebP format for all processed images. This significantly reduces file sizes while maintaining visual quality, especially for images with transparency.
• Made WebP the default export format for new images.
• Moved quality control to per-image settings via ImageSettings component, allowing individual quality adjustment per image instead of global quality.
• Quality slider now appears in ImageSettings panel with range 50-100% and default 80%.
• Download button now performs real format conversion using Canvas, converting images to their configured format (WebP, PNG, or JPEG) at download time with their individual quality settings.
• Added individual download button in ImageSettings panel to download specific image with its configured format and quality.
• Download button tooltip updated to "Download all images" to reflect mass download functionality.
• ImageSettings now available in all tool tabs (Remover, Colorizer, Upscaler) for consistent per-image configuration.
• Image processing now uses standard quality (85%) independent of per-image settings. Settings (format and quality) are only applied at download time for maximum flexibility.

🐛 Fixed

• Fixed bug where changing image settings in Colorizer or Upscaler tabs would cause the processed result to disappear from the canvas. This was caused by a useEffect dependency on settings array that incorrectly cleared resultData.

v1.1.0 - Canvas Improvements & Colorizer Tool

[1.1.0] - Canvas Improvements & Colorizer

✨ Added

• Image colorization feature using AI models (DeOldify)
• Custom zoom controls for canvas (zoom in, zoom out, reset)
• Zoom level indicator showing current zoom percentage
• Keyboard shortcut support: Ctrl/Cmd + mouse wheel for zoom
• Floating version button on canvas that displays changelog dialog
• Changelog dialog with GitHub link button
• Dynamic CHANGELOG loading from public folder using fetch
• Progressive Web App (PWA) support with Serwist and Turbopack
• PWA install button on canvas (shows when app can be installed)
• Service worker for offline functionality
• Web app manifest with proper icon configuration

🔄 Changed

• Removed react-infinite-viewer dependency
• Implemented custom canvas zoom using CSS transforms
• Canvas now centered with flexbox instead of InfiniteViewer
• Zoom range limited between 0.5x and 3x with 0.2x increments
• Canvas now adapts to image aspect ratio instead of forcing rectangular shape
• Canvas dimensions limited to max 800x600 for very large images

🧠 Improved

• Cross-platform zoom support (Ctrl for Windows/Linux, Cmd for Mac)
• Smooth zoom transitions with CSS animations
• Better canvas performance without external viewer library
• Smooth transitions for canvas dimension changes
• React-markdown configured with custom components for better changelog rendering

What's Changed

• Implement Custom Upscaler, Image Colorizer and SEO Improvements by @yossTheDev in #11
• Advanced SEO Implementation, Performance & Mobile Access Control by @yossTheDev in #12
• Fix colorizer logic and support multiple image sizes by @yossTheDev in #13

Full Changelog: v1.0.0...v1.1.0

v1.0.0 - First Stable Release

[1.0.0] - Stable Release

⚠️ From this version forward, the changelog begins official tracking.
Removerized has gone through multiple stages prior to this release — including early open-source iterations and an experimental SaaS direction.
Version 1.0.0 represents the first stable definition of the project as a local-first, open-source AI image toolkit.

---

✨ Added

• Background removal using multiple AI models (ONNX Runtime Web)
• Image upscaling feature
• Batch processing system
• Model selection interface (user can choose processing model)
• IndexedDB model caching for offline reuse
• Fully client-side processing (no backend)
• Offline support after initial load
• Interactive before/after comparison slider
• Queue system with preview thumbnails
• Export options (PNG with transparency)
• Modern UI with shadcn/ui + Tailwind CSS

---

🧠 Improved

• Performance optimizations for ONNX inference in the browser
• Memory handling for large image processing
• UX improvements in processing flow and controls
• Model loading and caching strategy
• Overall UI consistency and responsiveness

---

🔄 Changed

• Project direction finalized as open-source, local-first tool
• Removed dependency on external background removal libraries
• Reworked internal processing pipeline to support multiple models
• Improved architecture for scalability and future features

---

❌ Removed

• Legacy implementation based on external libraries
• Previous SaaS-oriented structure and constraints
• Any server-side dependencies or processing

---

🚀 Notes

This release marks a key milestone:

• Stable foundation
• Clear product vision
• Ready for expansion (colorization, restoration, captioning, PWA, extension)

---

Full Changelog: https://github.com/yossTheDev/removerized/commits/v1.0.0