The four voices

Pick a voice in Settings (⚙) under VOICE CHARACTER. Each one draws from a different era of speech synthesis and is applied to every word you hear — bundled or live-typed — and cached separately, so switching is instant after the first play.

Clean system speech with no retro processing — for when you just want clear pronunciation.

Deep-baritone monotone with a slow, deliberate cadence. Fixed-pitch formant synthesis at ~65 Hz, time-stretched 1.6×. A nod to late-60s sci-fi computer voices.

The original Speak & Spell character — a 10-pole LPC vocoder driven by the actual TMS5220 chirp. Buzzy and unmistakable. Cleaned up in 1.2.6 for even loudness and no glitches: every word sits at a consistent level, with no glitch on hard sounds.

Fixed-pitch impulse synthesis through formants extracted from your input, modelled on the Klatt-style voice that defined Hawking's computer. Deliberate and emphatically computery.

Any word beyond the 500+ bundled words falls back to on-device speech, routed through your chosen voice and cached to disk. Nothing leaves your device.

SpellSynth speaks French too. Pick FRANÇAIS under VOICE LANGUAGE in Settings and the voice, the words and the game screens all switch over. It's built after the 1981 French edition of the spelling toy: ÉPELLE, MYSTÈRE and DIS-LE with 360+ French words, plus an apostrophe key for words like L'ÉCOLE.

Accents show on the display but are never typed — spell ÉCOLE as E-C-O-L-E, just like the original. Every vintage voice works in both languages. Settings and the tour stay in English.