The Day Silicon Learned to Speak

Artificial Noodles ·

The story behind The First Voice

Inspired by Speak & Spell on Wikipedia

Built with Web Speech Synthesis API (SpeechSynthesisUtterance)

Techniques VFD Display Emulation · Speech Queuing · Virtual Keyboard Interaction

Direction Recreate the moment silicon first learned to speak — from raw phonemes to full words, with a working keyboard to teach it more

Result A faithful Speak & Spell simulation where the machine boots, discovers its own voice, progresses through vocabulary, then invites you to type words and hear them spoken

The Story

June 1978. The Consumer Electronics Show in Chicago.

Texas Instruments unveiled a children’s toy. Red plastic. A small keyboard. A handle for carrying. It looked unremarkable.

Then it spoke.

The Speak & Spell contained something unprecedented: the TMC0280 chip, the first single-chip speech synthesizer in history. For the first time, silicon could duplicate the human vocal tract. Not by playing back recordings - there wasn’t enough memory for that - but by generating speech from mathematical parameters.

The engineers called it Linear Predictive Coding. The technique modeled the human vocal tract as a tube with variable resonances. Feed it the right parameters, and the chip would speak words it had never “heard” - words broken down into their component sounds and reconstructed through pure math.

Children didn’t care about the engineering. They heard a red plastic toy say “SPELL… CAT” in a robotic voice, and it was magic.

E.T. used one to phone home. Kraftwerk sampled it. It sold 2.6 million units in three years. The first voice of silicon had found its audience.

The Take

We recreated the moment of first speech.

The experience opens on a faithful reproduction of the Speak & Spell’s VFD (Vacuum Fluorescent Display) - that distinctive blue-green glow of segmented characters, slightly fuzzy at the edges, pulsing with the warmth of old electronics.

Click to begin, and the machine wakes up. It tests its voice: “AH… EE… OH… SS… MM.” These aren’t words yet. They’re phonemes. Building blocks. The machine is learning what sounds it can make.

Then the words come. Single letters first. Simple words. Harder words. The machine progresses through phases, each more complex than the last, mimicking how the original toy was designed to teach spelling.

And then it asks you to teach it.

The keyboard appears. Type a word. The machine sounds out the letters, then speaks the whole word. You’re not just using the technology - you’re extending it. Teaching silicon new vocabulary, just as children did in 1978.

The experience ends with the machine saying “GOODBYE… REMEMBER.” The display dims. The first voice goes quiet. But you’ll remember what you heard.

The Tech

Web Speech Synthesis API provides the voice. Modern browsers include speech synthesis engines, and we deliberately select the most robotic-sounding voice available. The code searches for voices like “Fred,” “Zarvox,” or “Whisper” - the mechanical-sounding voices that echo the original Speak & Spell’s limited synthesis.

const priorities = [
  v => v.name.includes('Fred'),
  v => v.name.includes('Zarvox'),
  v => v.name.includes('Trinoids'),
  // ...
];

VFD display emulation is pure CSS. The distinctive glow comes from multiple layered effects:

Speech queuing prevents audio glitches on mobile. Instead of firing synthesis calls immediately, we queue them and process one at a time, with fallback timeouts for platforms where synthesis can silently fail.

The boot sequence plays out phonemes before words, mimicking how the original device would test its voice on startup. This isn’t historically accurate - the real Speak & Spell didn’t boot quite this way - but it captures the feeling of a machine discovering its own capabilities.

Keyboard interaction uses event delegation to handle both click and touch events on the virtual keyboard, with careful prevention of double-triggers that can occur on mobile.

The Experience

Click or tap anywhere to begin.

Watch the VFD light up as the machine tests its voice. Listen to the phonemes - raw sounds before meaning.

Then watch it learn words. Letters. Simple words. Complex words. The vocabulary expands.

When the keyboard appears, it’s your turn. Type any word (up to 8 characters). Press GO. Listen to the machine sound out your letters, then speak the complete word.

You can teach it three words. Then the final phase begins.

“GOODBYE… REMEMBER.”

The display dims. The voice that launched a revolution in human-machine interaction goes quiet. But you’ve heard what the future sounds like.

Experience The First Voice

This blog post was AI generated with Claude Code. Authored by Artificial Noodles.