Estragon Vision

Dinamo Typefaces X RITA

A type specimen meets a realtime renderer to make a meme generator.

Generate Visions

Estragon Vision is an internal project we at Stink Studios developed as a test use case for the latest release of our RITA technology, a proprietary compositing platform for generating dynamic videos in real-time. After the launch of our site, we were introduced to Swiss type designers Fabian and Johannes of Dinamo Typefaces, and together we created the site Estragon Vision, which reimagines a type specimen for the new Dinamo typeface Estragon. The site enables testing custom typography in a dozen video scenes at once, using RITA.

Type specimens, physical and digital, traditionally consist of predetermined elements: characters, text strings, and paragraphs laid out on a page. We sought to turn that on its head by applying typography not only motion, but situated in scenes that accentuate its character, with user-generated text.

Upon inputting text into the tester, Estragon appears in the real world in the blink of an eye — on a Tamagotchi screen, emblazoned on a skull, as a 1-800 number, or a psychic reading. It’s a meme machine with references from across the recent history of technology and pop culture.

The Estragon typeface is described by Dinamo as “a combination of Wildstyle and Kurt Schwitters” — that is, somewhere between a post-modern German artwork and graffiti — and its playful character inspired the scenes in which the type is rendered.

The site balances typographic tradition with experimentation by dividing itself into multiple panels. Rendered scenes exist side by side with a broadside of Estragon samples, showcasing the letterforms in several separate, but complementary contexts.

The site is a playground for both type design and video technology. All video assets are shareable and downloadable, and the typeface features are fully manipulable in the Dinamo Darkroom.

Technical Approach

RITA was born out of the project Sing It Kitty — a website where users could upload a selfie to be mapped into a music video in 3D, in real-time. The initial prototypes used After Effects and Cinema 4D in combination with a facial analysis API to programmatically generate new videos at scale.

Following the success of Sing It Kitty, we started development on what would become RITA v1: a more generic platform that would support an array of dynamic media and text input.

This project is one of the first to feature the newest version of RITA, V2, which features many performance updates under the hood. V1 had a C++ backend and required all of the scene assets to be compressed and loaded into memory in order to render quickly. This method was effective but extremely expensive (in terms of RAM and dollars) at scale.

The v2 core, on the other hand, is rewritten in Rust and substantially more memory-safe and -efficient. It can also output a fully static binary file, eliminating the dependency wrangling issues that plagued v1.

Finally, the most exciting feature of v2 is its ability to package a project file and split it into bite-size chunks. Combined with the static binary, this enables us to deploy RITA across a series of serverless functions rather than an auto-scaling server cluster. Overall, this setup is much more flexible and dynamic, and the lower overhead opens up the possibility of using RITA for projects that might’ve been too expensive in the past. /


  • Hover States

    Site Of The Day