86 lines (58 loc) · 2.64 KB

Changelog

All notable changes to SceneSplit will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

1.1.5 - 2026-01-16

Changed

Simplify release pipeline: Linux binary only, crates.io for all platforms
macOS users should use cargo install scenesplit

1.1.4 - 2026-01-10

Changed

Drop macOS Intel (x64) pre-built binaries (Intel users can use Rosetta or build from source)

1.1.3 - 2026-01-10

Fixed

Update CI to use macos-15 runner (macos-13 retired)

1.1.2 - 2026-01-08

Fixed

Fix cross-platform OpenCV compatibility (use cvt_color_def instead of cvt_color with AlgorithmHint)

1.1.1 - 2026-01-08

Fixed

Add missing LICENSE file for crates.io publishing

1.1.0 - 2026-01-07

Added

Metadata output: JSON sidecar files with timestamp, segment index, and frame number
Packaging improvements for crates.io distribution
GitHub Actions CI/CD pipeline
- Automated testing on push/PR
- Multi-platform release builds (macOS ARM64, macOS x64, Linux x64)
- Automated GitHub releases on tag push
- Automated crates.io publishing

Fixed

Resolved Clippy warnings for cleaner CI builds
Fixed rustfmt formatting issues
Added llvm dependency for opencv-rs compilation in CI

1.0.0 - 2026-01-01

Added

Initial MVP release
Video ingestion and offline decoding (no network access required)
Semantic embedding using MobileNetV3 ONNX model
Cosine similarity-based scene change detection
Deterministic frame selection (one representative frame per segment)
CLI interface with two primary controls:
- --detail flag: key (default), summary, or all
- --quality flag: fast, balanced (default), or best
Auto-download of ONNX model on first run
Support for common video formats via OpenCV
PNG output with timeline-ordered naming

Technical Details

Built with Rust for performance and safety
Uses ort (ONNX Runtime) for ML inference
OpenCV for video decoding
Fully offline operation after initial model download