PixKeep Lab

Browsers all claim to support the same image APIs. These experiments measure what they actually do — on your device, with your photos, nothing uploaded. Every experiment is open source and reproducible.

🧪 Native vs WASM image encoder benchmark

How good is the image encoder built into your browser? canvas.toBlob('image/jpeg' | 'image/webp' | 'image/avif', quality) at six quality levels, measured against WASM reference encoders (MozJPEG / libwebp / @jsquash-avif): byte sizes, encode latency and DSSIM quality scores, on your photos or a set of shared reference images.

Known catches so far: Safari ignores the WebP quality parameter entirely; no stable browser honours toBlob('image/avif') on the main thread (they silently hand you a PNG). Run it on yours:

Run the benchmark →

🔭 Coming soon

Format quality mapping (measured AVIF↔JPG↔WebP perceived-quality equivalents) and the DSSIM methodology behind them. Follow the repo or bring your own question to Discussions.

Why this exists

PixKeep builds in-browser image tools — everything client-side, nothing uploaded. That means living with the sharp edges of browser image encoders every day. Lab is where those edges get measured in the open, so anyone building on canvas, WASM codecs or the web platform can check the numbers instead of trusting the docs.