1.2.1 — PHRASE mode polish. • New GAP knob on PHRASE — live-tune the silence between words from 50ms to 500ms while you listen. Default 100ms (down from 150ms in 1.2.0) for tighter delivery. • Smoother per-word playback. Words now stitch through a single audio path, removing a small click some users heard on word boundaries. • Cleaner starting screen — PHRASE opens to an empty canvas. Type your own phrase via KEYPAD or TYPE; the RECENT row remembers everything you've played this session.

1.2.0 — multi-word phrases. Type a whole phrase, hear it spoken. The headline: a new fifth game mode called PHRASE. Type anything more than a single word — "GOOD MORNING", "I LOVE YOU", "MARY HAD A LITTLE LAMB" — and SpellSynth speaks the whole phrase through whichever vintage voice you've picked. The VFD cycles one word at a time as the audio progresses, so reading along is natural. Two ways to type: • KEYPAD — the familiar yellow A-Z grid, now with a SPACE key. Each letter still plays on tap; SPACE silently separates words. Works with MIDI keyboards too. • TYPE — for paste, dictation, or the iOS keyboard. Tap the field, type a phrase, hit SAY. Twelve preset phrases come baked in — greetings ("HELLO WORLD", "GOOD MORNING"), early-reading sentences ("THE CAT IS HERE", "I CAN READ"), old-computer-movie callouts ("SYSTEM ONLINE", "GAME OVER"). Tap any chip to play it. Phrases you've played save in a RECENT row for the rest of the session. All vintage voices work on phrases, though each has its own character at sentence length. 1986 DECTALK STYLE sounds the most natural — DECtalk itself was built for sentences. 1978 LPC keeps the Speak & Spell buzz but smooths word boundaries (the TI chip was tuned for sub-second utterances). 1968 ODYSSEY's deliberately slow 1.6× cadence makes phrases feel measured. In the AUv3 plugin: the voice mode (RAW vs 1978 LPC) used to be a knob buried in the MIDI panel. It's now a dedicated VOICE segmented picker, sitting above PITCH where you'd actually look for it. Saved presets and host automation still resolve to the same parameter — nothing breaks. A small note from me: SpellSynth is still a one-person side project I keep poking at on weekends. If you've got an idea, hit a bug, or just want to say hi, zhiii0x@gmail.com is the fastest way to reach me — replies go to my own inbox, not a queue. And if SpellSynth has made you smile even once, a quick App Store review goes a long way for a one-person shop. — Zhi

1.1.8 — a third vintage voice, and a long-overdue layout fix for tiny iPhones. The headline: a new voice character called 1968 ODYSSEY, sitting alongside 1978 LPC and 1986 DECTALK STYLE. Calm, deep-baritone monotone with a deliberately slow cadence — an aesthetic homage to the late-60s sci-fi AI voices that defined how a generation of audiences expected computers to talk. Same fixed-pitch formant synthesis as 1986 DECTALK STYLE underneath, but pitched a third lower (~65 Hz) and time-stretched 1.6× so each word lands slower. The voice picker grew into a 2×2 grid in Settings to make room for the fourth character (DEFAULT / 1968 ODYSSEY / 1978 LPC / 1986 DECTALK STYLE). Cells got bigger labels too — easier to read at arm's length and less squashed against the bottom of the sheet. Layout fix: • iPhone 13 mini with Display Zoom turned on (320pt logical width) was clipping the keypad and the mode bar off the right edge of the screen. The VFD now shrinks to fit narrow canvases instead of dragging everything off-screen with it. Standard-mode mini and every other iPhone are visually unchanged. Other polish this round: • Settings sheet now opens at full height instead of half, so the voice picker isn't squashed against the bottom edge. • Onboarding font sizes bumped on the modes and voices pages — the cards were holding 10/11pt labels in space that could comfortably fit 13/16pt. A small note from me: SpellSynth is still a one-person side project I keep poking at on weekends. If you've got an idea, hit a bug, or just want to say hi, zhiii0x@gmail.com is the fastest way to reach me — replies go to my own inbox, not a queue. And if SpellSynth has made you smile even once, a quick App Store review goes a long way for a one-person shop. — Zhi

1.1.7 — a second vintage voice, plus a guided tour for new players. The headline: there's a new voice character called 1986 DECTALK STYLE, sitting next to the 1978 LPC voice that's been here since launch. It's a fixed-pitch, monotone formant synth — same family of tech that gave Hawking his iconic computer voice. Chunkier and more deliberate than the 1978 LPC's buzzy texture, and it works on every word — bundled or live-typed — because it re-synthesises whatever the speech engine produces. The voice picker has moved into Settings, where it makes more sense to live as a "set once" preference. Three-way row: • DEFAULT — clean system speech, no extra processing. • 1978 LPC — the original Speak & Spell character (TI TMS-style chirp). • 1986 DECTALK STYLE — fixed-pitch monotone, computer-voice feel. The standalone LPC button on the keypad is gone; whichever voice you pick now shows up as a small caption above the function row so you always know what's active. Other changes this round: • First-launch tutorial — a quick four-page tour of the modes and the three voices. Page three lets you tap each voice card to preview it in place before committing. Reach it again any time from Settings → REPLAY TUTORIAL. • The "Rate SpellSynth" prompt now fires on app opens (third, tenth, thirtieth) instead of after a spell win — felt closer to "you keep coming back" than "you spelled CAT correctly". The manual RATE row in Settings still works any time. • Smaller layout polish across the Settings sheet and the What's New sheet. A small note from me: SpellSynth is a side project I keep poking at on weekends. If you've got an idea, hit a bug, or just want to say hi, zhiii0x@gmail.com is the fastest way to reach me — replies go to my own inbox, not a queue. And if SpellSynth has made you smile even once, a quick App Store review goes a long way for a one-person shop. — Zhi

Version 1.1.5 — pronunciation polish and a tighter, more reliable layout across every iPhone and iPad size. • Free Play now speaks ANY word, not just the bundled ones. Type a word, tap SAY, and it's pronounced on the fly via the system speech engine — then cached so the second tap is instant. Fixes a bug where unknown words produced silence. • Short words like HI and OK read as words instead of being spelled out as initialisms (H. I., O. K.). Fixed by lowercasing input before it reaches the synth — the engine's acronym heuristic was triggering on short uppercase strings. • Toggling the 1978 VOICE button no longer shifts the display. The VFD stays put through the toggle and through any LPC-related re-render, so the keypad doesn't reflow under your finger. • Bigger, easier-to-read VFD on iPhone, with the cells locked to hardware-spec proportions so a tight layout can never squash them down to a sliver. • 1978 VOICE is ON by default for new installs — the brand voice IS the brand. Your existing preference is preserved if you've already toggled it off. • Layout overhaul: iPad portrait now uses the compact "handheld" layout with the mascot floating below, since the two-column treatment was clipping on iPad 11" / iPad Air 11". Landscape iPad still gets the wide two-column layout. • Tighter spacing on small iPhones (iPhone 16e and similar) so the mascot stays fully on screen instead of being cut off at the bottom edge. • First-launch "What's New" sheet to surface each release's changes, and a small version number in the bottom-right corner for easier feedback / TestFlight reports.

Version 1.1.0 — the original Speak & Spell voice arrives, plus a much bigger MIDI surface for AUv3 hosts and availability on Mac and Apple Vision Pro. • NEW: LPC voice mode. Tap LPC in the brand header to re-synthesise every letter and word through a 10-pole LPC vocoder driven by the actual TMS5220 chirp ROM — the same speech algorithm the 1978 toy used. The AUv3 plugin gets its own LPC knob in the MIDI panel. Both versions cache so toggling is instant after first play, and your choice persists across launches. • Sustain pedal (CC 64) now defers note release until you lift the pedal — proper MIDI sustain behaviour for hold-and-release phrasing. Channel Pressure (aftertouch) is routable to any parameter via MIDI Learn, tagged "AT" in the bind list so you can tell aftertouch and CC bindings apart at a glance. • Velocity curves: Soft / Linear / Hard, selectable in the AU's MIDI panel. Pitch-bend range becomes a knob too — anywhere from ±0 (locked) to ±12 semitones (a full octave bend). • MIDI Learn list grows from 5 to 14 controllable targets: pitch, speed, crush, cutoff, resonance, attack, release, three LFO destinations, stutter on/off, stutter rate, reverse, and the Rate knob — every performance parameter, mappable. • External MIDI keyboards now drive the standalone app's keypad. Connect any USB, Bluetooth, or Network MIDI controller and notes C4–C#6 trigger letters A–Z just like tapping the keys on screen. Works in every game mode. • Now available on Apple Silicon Mac and Apple Vision Pro via Designed for iPad. Same iOS app, no separate download. • Reliability under the hood: the AUv3 voice bank is lock-free across the audio thread, runtime caches are bounded so long DAW sessions don't balloon memory, and corrupted user presets log a warning rather than silently vanish from the picker.

Version 1.0.5 — answering the first round of AUv3 reviewer feedback. • The AUv3 plugin now speaks whole words, not just single letters. A new mode switch picks between LIBRARY mode (type any word into the text field; 500+ bundled retro samples play back as one hit, unknown words gracefully fall back to letter-by-letter spelling) and LETTERS mode (a full yellow A–Z keypad just like the original toy). A "View list" link in the plugin opens the full word catalog in your browser. • Fixed: AUv3 voice was playing a full octave higher than intended. The bundled samples are 24 kHz, but the plugin was reading them at the host's sample rate — which meant every sound came out one octave up. Now runs at native pitch on any host rate, with a regression test locking the contract so it can't drift again. • New Rate knob in the TONE panel drops the effective sample rate toward 8 kHz to emulate the stepped DAC character of a 1978 speech chip. Pair it with Crush for very vintage territory. • New "Classic 1978" factory preset stacks crush, rate, and cutoff into the crunchiest voice in the bank — the closest we can get to the TI0280 speech chip Apple won't let us ship directly. • Recently typed words save as one-tap chips under the text field for quick recall.