modern-format-boost

Modern Format Boost is a media optimization tool written in Rust that provides separate pipelines for static images and video/animated media. It emphasizes evidence-driven optimization, ensuring that files are only delivered after passing strict decode, quality, and integrity gates. Key capabilities include: - **Static Image Pipeline (`img`)**: - Supports conversion to JXL and AVIF. - Features a reversible JPEG to JXL path that ensures byte-identical reconstruction. - Handles lossless conversion for PNG, TIFF, BMP, and other raster formats. - Implements "AVIF Meme Mode" for bounded size/quality searches. - Preserves EXIF, XMP, ICC profiles, and macOS-specific metadata (xattrs, Finder tags). - **Video Pipeline (`vid`)**: - Transcodes video and animated rasters (GIF, WebP) to HEVC or AV1. - Employs a three-phase saturation search: GPU coarse search, CPU fine-tuning, and a 3D quality gate using VMAF, CAMBI, and PSNR. - Supports HDR10+ dynamic metadata and Dolby Vision RPU extraction/injection. - **Technical Features**: - Content-based format identification using magic bytes rather than file extensions. - BLAKE3-fingerprinted analysis cache for efficiency. - Local-first architecture with no telemetry or cloud uploads. - A native macOS AppKit application for drag-and-drop batch processing. - Fail-closed design that retains original source files if quality or size policies are not met.