Gianni Rosato

Introducing Iris-AVIF, a better AVIF encoder & the best image encoder in the world: halide.cx/blog/introdu... Iris-AVIF offers up to >21% better compression than libaom, the reference AV1 encoder with image-specific tuning. #av1 #avif #jxl

Halide Compression's open-source, memory-safe WebP decoder is here: halide.cx/blog/wpd We <3 open source! #webp #rust

⚡Featured posts this week: ✨ New ad metrics in CrUX: @alextcone.bsky.social ✨ The case against JPEG XL: @giannirosato.com ✨ Why Spotify Is Not Using Bayesian A/B Testing: Mattias Frånberg, Mårten Schultzberg ✨ International SEO: @polemicdigital.com 🙏 Special thanks to the authors! #webperf

The case against JPEG XL Investigating JPEG XL's place as a Web image codec, by @giannirosato.com, via @webperfreport.bsky.social giannirosato.com/blog/post/ca... #JPEGXL #JXL #ImagePerf #WebPerf

This is a great article. I've always been sus about the hype around JPEG XL, as it never stacked up when I actually tested it with web use-cases. giannirosato.com/blog/post/ca...

New Halide blog post on some of the weaknesses of perceptual metrics: halide.cx/blog/metric-... These lessons are helping us develop our next encoder, Aperture. #compression

It's really weird that the response to JPEG XL shipping is "yesss fuck WebP!" Like, we've had AVIF for years, and it outperforms JPEG XL in almost all web use-cases.

Hey everyone! fmetrics, a fast open-source perceptual metrics toolkit that runs on the CPU, is here: github.com/halidecx/fme... fmetrics supports IW-SSIM, MS-SSIM, CVVDP, SSIMULACRA2, & Butteraugli. See the blog post (halide.cx/blog/fmetrics) for more info. All code is Apache 2.0. Enjoy!

Making Software: Image Compression www.makingsoftware.com/chapters/ima... Cool to see WebP addressed here too!

I've been improving my image compression project on and off for the last six months or so. Today I got it to basically where I want it feature-wise, so I've put out an alpha release and written a quick little blog post about it: www.rachelplusplus.me.uk/blog/2026/06...

fcvvdp 0.2.0! github.com/halidecx/fcv... Highlights: ~9-12% speed boost for single-threaded, task threading (220% faster than the reference implementation with 8 threads on my 13700k), FFmpeg support via patch. Check it out and lmk what you think!

I've put together an updated version of the Sponza scene with uncompressed PNG and compressed AVIF textures. I wrote about the process and compared the results against KTX. www.ludicon.com/castano/blog... #webgpu #web3d #sparkjs

New Halide blog post on post-processing for chroma as it is relevant to compression efficiency: halide.cx/blog/chroma-... Seems like you can get a roughly 2% efficiency improvement with 4:2:0 decoder output just with smarter post-processing for chroma. Pretty neat! #avif #webp #jpeg

the 2nd part of my dithering visual article is finally out! 🔗 visualrambling.space/dithering-pa... this one mainly explores the threshold map and how it generates those unique visual patterns hope you enjoy this as much as I enjoyed making it! made with #threejs & #animejs

SVT-AV1 4.0.0 is here! Changelog: gitlab.com/AOMediaCodec... SVT-AV1-PSY's porting efforts are fully complete, and this release also features huge AVIF improvements. Maybe the most exciting release to date by my standards – SVT-AV1 is finally a real perceptual encoder! #av1 #avif #compression

Happy Holidays! Introducing fcvvdp, a faster CVVDP implementation: github.com/halidecx/fcv... Compared to the reference implementation running on the CPU, fcvvdp is faster (~15%) despite being single-threaded. It sees stronger wins with user time (up to ~15x) and memory usage (>90% less memory).

SVT-AV1-HDR 3.1.3 "Cyclonus" is out! Including all-new features from SVT-AV1, plus code cleanups and stability fixes. Get the newest version of this all-purpose AV1 video encoder! ffmpeg and HandBrake builds are available by clicking the link below: github.com/juliobbv-p/s...

Enhancing HDR on Instagram for iOS With Dolby Vision: https://engineering.fb.com/2025/11/17/ios/enhancing-hdr-on-instagram-for-ios-with-dolby-vision/ #av1 #vp9 #hdr

Introducing oavif: faster target quality image compression giannirosato.com/blog/post/oa... oavif can be up to 63% faster than traditional target quality encoders. Learn how in the attached blog post! #av1 #avif #compression

fssimu2 has its first release! github.com/gianni-rosat... fssimu2 is a fast SSIMULACRA2 implementation in Zig, with support for distortion map output. It is ~23% faster than the reference implementation at ~40% less memory usage, while being 99.97% accurate. Give it a look if you're interested!

New Halide blog post! halide.cx/blog/consist... Quality and speed define an image encoder's compression performance. Consistency is a close third, and easily overlooked in image encoder design. What value does it provide, and how can we measure it? #avif #webp #jxl #jpegxl #compression

The Halide Compression website (halide.cx) now has an Atom feed – just paste the URL into your favorite RSS reader and you should be good to go!

New blog post about my experiences using spark.js with three.js: www.ludicon.com/castano/blog... Re-posts and feedback appreciated!

Anyone using rss/atom readers can now type `giannirosato.com` into their reader to get my website's atom feed

I got to interview Julio Barba from Google for the Halide blog: halide.cx/blog/julio-b... Go check it out if you're interested in what the next decade of multimedia tech might look like!

Completely redesigned my website with Zola (www.getzola.org) – check it out at giannirosato.com Let me know if any links are broken if you link to my site – it should be fully backwards compatible with my old link hierarchy

The “AV1 is a disaster” myth Here’s what’s actually true: - SVT-AV1 is fast, fully multithreaded and packed with SIMD. - It scales beautifully on modern CPUs. - It beats x264, x265, and VP9 in bitrate vs runtime. AV1 on CPU isn’t a disaster. It’s quietly winning.

Tune 3: Still Picture is coming to SVT-AV1: https://gitlab.com/AOMediaCodec/SVT-AV1/-/merge_requests/2489#note_2704093242 Formerly Tune 4 in SVT-AV1-PSY. Thank you @juliobbv.bsky.social !

Exciting #avif news – libaom’s tune iq (derived from SVT-AV1-PSY’s Tune 4) is used across The Guardian’s website now. They say they serve mostly AVIF, with JPEG XL shipped when supported. Very exciting stuff! Thanks @juliobbv.bsky.social!

New post! After my previous post about image compression, a few people asked me to try different encoder settings. So I went ahead and tested a couple of them :) www.rachelplusplus.me.uk/blog/2025/07...

Improving AVIF in Open Source via the Halide blog halide.cx/blog/improvi... I wanted to write this to explain some of what went into Tune 4/Tune IQ, & help outline some of my reasoning for starting Halide Compression in the first place. I hope you enjoy! #av1 #avif #encoding

AV1 @ Scale: Film Grain Synthesis, The Awakening netflixtechblog.com/av1-scale-fi...

The world’s largest digital camera — which will create an unparalleled map of the night sky — is at the heart of the new Vera C. Rubin Observatory in Chile. Here’s how it works. nyti.ms/4n8OpgP

"I designed my own ridiculously fast game streaming video codec – PyroWave" themaister.net/blog/2025/06... Cool to see github.com/psy-ex/metrics in use in the wild, I'm glad it was useful!

New blog post: I've spent the last couple of months building a tool to compare different image encoders, and wrote an article about some of the things you need to consider to get reliable results. I also talk a bit about how JPEG-XL is looking promising! www.rachelplusplus.me.uk/blog/2025/06...

Help restore the tax deduction for software dev in the US (Section 174) news.ycombinator.com/item?id=4422... If you live in the US and you've ever used a piece of software, you're probably interested in this. There's a sign-able letter attached.

May have bitten off quite a lot with this next blog post, but the goal is in sight! Basically built an encoder comparison tool from scratch, to really nail down the details in my own mind. Should have some interesting things to say soon - think of it as a "yes, and" to Gianni's excellent post:

There's simply no way

Lil' teaser for an upcoming blog post, building on my last post (www.rachelplusplus.me.uk/blog/2025/03...): This graph compares compressing an image at a single fixed resolution (dotted curve) vs. encoding at multiple resolutions, netflix style (solid curve):