RE: https://typo.social/@triple/117419027672718120
I'm flattered to having made it to this list. Needless to say, my work on HarfBuzz in the past twenty years has been collaboration with dozens of colleagues, and generously funded by Google, to which I'm grateful.
The 'boring-expansion' set of proposals that I made back in 2022, with the help of Google Fonts and industry colleagues, have finally become part of the new ISO Open Font Format 5th edition international font format standard. It _only_ took four years...
The proposals include the `avar` version 2 table for more expressive variable font designspaces, variable-composites, as well as fonts with more than 65 thousand glyphs.
PDF can be downloaded for free, albeit after a checkout process and watermarked to the individual.
Update on the Slug algorithm Open Source implementation: I absorbed GLyphy into a new HarfBuzz module called hb-gpu.
Check out the web demo. Interact with it with the mouse. Or press space to animation rotation:
https://behdad.org/hb-gpu-demo
Discuss hb-gpu at:
RE: https://typo.social/@behdad/116247460216755113
And... I have a patchset to change GLyphy to use the Slug algorithm & shaders. Runs >2.5x faster than GLyphy, without any of the rendering artifacts.
Great news today for GPU-accelerated text rendering:
"""
I was granted a patent for the Slug algorithm in 2019, and I legally have exclusive rights to it until the year 2038. But I think that’s too long. The patent has already served its purpose well, and I believe that holding on to it any longer benefits nobody. Therefore, effective today, I am permanently and irrevocably dedicating the Slug patent to the public domain. ...
To aid in implementations of the Slug algorithm, reference vertex and pixel shaders based on the actual code used in the Slug Library have been posted in a new GitHub repository and made available under the MIT license. ...
"""
Thank you, @EricLengyel!
OpenType Variable Fonts tech was released on September 14, 2016. Many downplayed its adoption in the following years, or even called it a failure. Early on, I shared my vision that in ten years time we will see variable-fonts just be the standard font format in use, the same way that scalable fonts replaced bitmap fonts.
Ten years have passed, and at least to me, when I look around, it seems like most new font families are designed variable-first. Is that accurate to say? That sounds like widespread adoption to me.
This year we should see the complete standardization of two new variable-font technology additions: avar2 and variable-composites (VARC table). I am very excited about both technologies and seeing how they roll out and affect how var fonts are designed.
HarfBuzz 13 ships with the new `hb-raster` library. You might ask why yet another Open Source rasterizer when there are many high-quality ones available.
`hb-raster` fills two gaps in the Open Source font rasterization scene:
1. Many graphics-oriented frameworks & mobile apps don't want to include the entirety of FreeType to get gain text rendering. Many resort to `stb_truetype`, which doesn't support variable fonts, and crashes on bad font data. `hb-raster` addresses both of these issues.
2. Before `hb-raster`, rendering COLRv1 fonts was not possible without whipping up your own renderer, or pull in Cairo or Skia, both of which are otherwise unnecessary for many clients. `hb-raster` renders COLRv1 and SVG color fonts natively, & soon bitmap-based color-fonts too.
Calling on all downstream projects (browsers, Linux distros, etc) to share their thoughts on AI contribution policy for HarfBuzz:
RE: https://typo.social/@fontra/116183028572020334
This is quite an interesting and clever way to use #HarfBuzz to do shaping on the fly.
RE: https://typo.social/@behdad/116172838540880597
I'll spare you my rambling thoughts. If someone wants to fork HarfBuzz 12.3.2, go ahead, that's before AI "slop". I let @khaled, the HarfBuzz maintainer, decide whether AI-assisted code is allowed in the upstream HarfBuzz or not.
As for code quality, I'm ultimately the only one responsible for code that I push out.
Finally, if this technology is good for Linus, Guido, and Knuth, among others, I'm not gonna ignore it and let my projects and skills become irrelevant and obsolete.
I respectfully ask that you refrain from your urge to leave a 🤮 or equivalent in the comments. Thanks.
RE: https://typo.social/@behdad/116172838540880597
Wow, responses to this caught new off-guard. I'll sleep on it before responding.
HarfBuzz 13.0.0 released with new experimental features: `hb-vector` for vector output of glyph outlines to SVG and `hb-raster` for rasterizing glyphs to A8 / BGRA32 images, among other features.
I would like to welcome our new contributors: Claude & Codex, without whom these new features would not have been possible.
Do you build avar2 variable fonts? I would appreciate your help testing the fonttools partial-instancing (aka designspace-limiting) PR at https://github.com/fonttools/fonttools/pull/4045
I have been doing a lot of AI coding with OpenAI Codex and Claude AI coding agents, and it has *completely* transformed how I work. I wrote my thoughts briefly on X:
I'm adding bunch of new functionality to the HarfBuzz cmdline tools, thanks to my codex / gemini AI coding interns.
I added `--ink` and `--logical` for choosing how to box the output. The default now unions the two. No cutoff glyphs by default anymore.
Next, I added `--stroke` and `--foreground=COLOR,...` for rotating rendering glyphs using a rotating palette. Particularly useful for Arabic and Indic font design.
Included is a `--rainbow` shortcut, with an Easter egg hidden `--lgbtq` alias 🏳️🌈.
welp. #reddit
A wave of manic energy in December had me put together a long deck called "HarfBuzz at 20! " , celebrating 20 years of HarfBuzz. 🎂
I designed the deck to be presented at the #WebEnginesHackfest later this year. Then reality hit that I cannot present this deck in any sane amount of time.
Inspired by all the great presentations coming out of #FOSDEM today, I decided that instead of tossing the deck out, I just put it out here to be read by the curious. I will present a highly condensed version at the hackfest in June.
Let me know what you think. 🙏
https://docs.google.com/presentation/d/1o9Exz1c-Lr-dJjA8dcBn_Vl_Y37cupmFzmclMjBE_Bc/view
Here's three graphs, showing the cost-per-character in fonts for three datasets of CJK fonts, built with successively more number of characters from the source, in both a flat OpenType font and one with the new `VARC` table, showing massive file-size saving potential of the new technology, specially in the case of Hangul.
I have an update about experimental HarfRust performance that I like to share:
To beat HarfBuzz with HarfRust
https://docs.google.com/document/d/1LjYFjZj8Kw8zyhqsfZg0_VgzHf2zhYUsksPCkZL7GJI/edit
TL;DR:
Previous performance comparisons of HarfRust to HarfBuzz were flawed in that the HarfBuzz code tested had assert statements enabled. This skewed the numbers in favor of HarfRust. Moreover, since the last measurements in September 2025, both HarfBuzz & HarfRust got faster, but at a rate that made HarfRust fall further behind HarfBuzz in comparison.
In this work, proof-of-concept unsafe patch to font-codegen & read-fonts as well as removing bounds-checking from HarfRust buffer type show massive speedup benchmarks: from being 20% to 80% slower, down to being less than 10% slower.
HarfBuzz 12.3.0 is out, faster than ever:
- AAT shaping: 12% faster, LucidaGrande benchmark.
- OpenType shaping: 20% faster, NotoNastaliqUrdu benchmark.
- Drawing mega variable-fonts: 30% faster, GoogleSansFlex benchmark.
- Drawing VARC fonts: 5% faster, varc-hanzi benchmark.