Iris-WebP, a new WebP encoder

[Gianni Rosato]

Hi WebP people!

I’m Gianni, a software engineer interested in compression. I started the SVT-AV1-PSY project for perceptual AV1 encoding, and became interested in image compression through working on AVIF. After seeing great results from that, I worked for Two Orioles (the team behind dav1d) on their video encoders for a while and eventually pivoted to working on image compression full-time via my own company, Halide Compression.

Since several people here serve WebP at scale, I thought it might be useful to share what I've been working on. Alongside consulting, for roughly a year now I’ve been developing a new state of the art WebP encoder called Iris-WebP (link: halide.cx/iris). My focus has been on making sure the encoder is extremely fast and perceptually excellent. According to the SSIMULACRA2 metric, Iris should be ~17% more efficient than libwebp while being faster (see the graphs for more info). Iris also wins on older metrics like PSNR and SSIM.

The encoder is proprietary, but if your company delivers WebP images at scale, maybe there is an opportunity for Iris to provide value to you. If you’re interested in testing or evaluating Iris, you have any questions, or you just want to talk about WebP, feel free to reach out!

Best,

Gianni

gia...@halide.cx

https://giannirosato.com/