Reconstructs lost JPEG detail for smoother, cleaner images with reduced compression artifacts.
Get running in less than 60 seconds:
# 1. Install
brew install Delnegend/tap/artefact-cli
# 2. Reconstruct an image
artefact-cli input.jpg -o output.png
# 3. Web browser (zero install)
# Visit https://artefact.delnegend.com to drop and compare images locally with WebAssembly/WebGPU.- Artifact reduction — Re-optimizes raw DCT coefficients with a regularized solver rather than making pixel-level guesses.
- High throughput — Multi-threaded execution with adaptive SIMD vectorization and optional WebGPU acceleration.
- Cross-platform — Runs as a single static native CLI or entirely client-side in the browser via WebAssembly (zero uploads).
- Flexible output — Ingests standard
.jpg/.jpegfiles and exports to PNG, WebP, TIFF, or BMP.
# Typical command for daily use
artefact-cli input.jpg -o output.webp --gpu -y| Option | Default | Description |
|---|---|---|
-o, --output <path> |
<input>.png |
Path for reconstructed output image |
-f, --format <fmt> |
auto |
Output format (auto, png, webp, tiff, bmp) |
-w, --weight <f32> |
0.3 |
Smoothness weight (higher = smoother transitions) |
-p, --pweight <f32> |
0.001 |
Fidelity weight (higher = closer to source JPEG) |
-g, --gpu |
false |
Enable GPU solver with automatic CPU fallback |
For all flags, per-channel tuning, and benchmark options, see CLI Reference.
flowchart LR
JPEG[JPEG Input] --> Decode[zune-jpeg fork]
Decode --> Solver["artefact-core (Scalar / SIMD / GPU)"]
Solver --> CLI[artefact-cli]
Solver --> WASM[artefact-wasm]
WASM --> Web[Web Frontend]
For component boundaries, internal pipelines, and design decisions, see Architecture Guide.
- CLI Reference — Command-line flags, tuning guide, and benchmark options.
- Architecture — Solver pipelines, data flow, and repository layout.
- Development — Prerequisites, builds,
just check, and profiling.
Dual-licensed under MIT or Apache 2.0.