Soundscribe

Your words become sound, and sound becomes ink.

What is it?

Write a short letter and Soundscribe reads it aloud then prints the waveform of that speech as a letter you can keep. Every letter looks and sounds a little different, because it's made from the shape of your words.

Where did you get this idea?

In 2010, Scottish artist Jen Cantwell made a piece called Letter Home. It imagined a migrated bird writing home, its voice rendered as waveforms on paper. It became viral on Tumblr years later.

Then 5 years ago, a developer who goes by @Amustache saw the piece online and built Cantwell: a Python CLI tool that converts text-to-speech audio into visual soundwave PDFs. He named it after her. That tool is the technical ancestor of everything here.

I went through a rabbithole on soundwaves one night and found Jen's art and Amustache's github, and I thought this project needed a place where people could also write their own letter. So I built soundscribe.

How it's built

I used Cantwell as a base, but ported the whole thing from Python to TypeScript, running fully client-side in Next.js. The core algorithm stayed the same, I just rewrote it so it runs in the browser instead of a server.

For the TTS I'm using my own voice that I cloned with Gradium, and it falls back on Kokoro when I run out of credits lol.

For the Three.js I used an open source repo from hoainho called img2threejs.

Open source and MIT licensed.

Source on GitHub About the author