Discover more from hrbrmstr's Daily Drop
Drop #345 (2023-10-03): Typography Tuesday
⚠️ Cloudflare Fonts; Hey, Computer, Make Me a Font; Tilt Fonts
Today, we don some tin foil hats as I go all “Alex Jones” on a new font CDN, look at a way to make fonts with “AI”, and then check out some wicked cool glyphs that add some dimensions to your typographic creations.
This is an AI-generated summary of today's Drop.
The first section criticizes Cloudflare Fonts for claiming to enhance privacy while still collecting user data, and advises readers to consider serving fonts from their own servers instead (source).
Sergey Tselovalnikov's post, “Hey, Computer, Make Me a Font,” details his journey in creating generative machine learning models to produce unique fonts, with a GitHub project available for those interested in trying it out.
Tilt is a family of three variable fonts inspired by dimensional lettering styles, offering designers control over the horizontal and vertical rotation of the tilting letterforms, with the project available on GitHub.
⚠️ Cloudflare Fonts
While I'm loathe to recommend using either Google Fonts directly, I'm even mroe loathe to recommend using Cloudflare for anything, including fonts, which they now serve.
Here's their adorable pitch:
When you load fonts from Google, your website initiates a data exchange with Google's servers. This means that your visitors' browsers send requests directly to Google. Consequently, Google has the potential to accumulate a range of data, including IP addresses, user agents (formatted descriptions of the browser and operating system), the referrer (the page on which the Google font is to be displayed) and how often each IP makes requests to Google. While Google states that they do not use this data for targeted advertising or set cookies, any time you can prevent sharing your end user’s personal data unnecessarily is a win for privacy.
With Cloudflare Fonts, you serve fonts directly from your own domain. This means no font requests are sent to third-party domains like Google, which some privacy regulators have found to be a problem in the past. Our pro-privacy approach means your end user’s IP address and other data are not sent to another domain. All that information stays within your control, within your domain. In addition, because Cloudflare Fonts eliminates data transmission to third-party servers like Google's, this can enhance your ability to comply with any potential data localization requirements.
Please do not be so naive as to believe this rot.
Sure, the connections go to your domain…via Cloudflare's servers. They (and, you) get all the telemetry instead of Google. You just choose between two evils.
I don't believe anything Cloudflare says when it comes to “privacy”. Their actions in other ways show they have little integrity or honor. Far too much of the internet goes through them already.
Having said that, if you do trust them, and you aren't willing to just serve fonts from your own servers, then you should consider switching to Cloudflare fonts.
Hey, Computer, Make Me a Font
“Hey, Computer, Make Me a Font” is a very well-crafted post by Sergey Tselovalnikov, who shares his journey of learning to build generative machine learning models from scratch and teaching a computer to create fonts in the process. This post provides an overview of the project and the techniques used to generate unique fonts using ML models.
To train the model, Tselovalnikov assembled a dataset of 71,000 distinct fonts. 60% of all fonts only had a vague category assigned to them, while 20% of fonts were accompanied by longer descriptions. The descriptions were condensed to a few keywords using GPT-3.5. Additionally, 15% of the fonts had prompts containing only the font's name, and the remaining 5% of the dataset had an empty textual description assigned to them. This ensured that the model was capable of generating fonts with various levels of information. (Aside: this painstaking training part is what makes “data science” feel like the kind of drudge work coal miners had to go through.)
Font embeddings were used to represent fonts in a high-dimensional space, allowing the model to understand the relationships between different fonts. This technique enabled the model to generate new fonts by interpolating between existing ones or by adding specific features to a given font.
If you're interested in generating your own font using this project, you can head to the GitHub project, clone it, and, perhaps, show some 💙 with a ⭐️. Then, download the weights from Huggingface and follow the instructions provided in the blog post.
Tilt (GH) (gDesign) is a family of three variable fonts inspired by dimensional lettering styles often found in storefronts. Designed by Andy Clymer in 2019, the idea for Tilt came from an ordinary neon sign in a New York City laundromat. The sign's unique appearance when viewed off-center sparked Clymer's interest in creating a variable font that could mimic this effect.
The Tilt family consists of three styles: Neon, Prism, and Warp. Neon is based on the original neon sign that inspired the design, while Prism references vacuum-formed plastic lettering found on theater marquees or cast bronze letters at bank entrances. Warp, on the other hand, takes inspiration from the modern trend of vinyl sticker lettering and the three-dimensional forms they create when peeling.
Tilt's variable font format allows designers to control the “Horizontal Rotation” (HROT) and “Vertical Rotation” (VROT) of the tilting letterforms, offering a unique approach to typography. The Neon style functions as a regular style in text, while Tilt Warp is heavy enough to work as a companion bold when necessary.
These are super cool, and being able to play with the rotation is just bonkers neat. I gotta find a reason to use these, soon.
If you do try your hand at an AI-generated font, def share it with the rest of us! ☮️
hrbrmstr's Daily Drop is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.