Diction

Voice-first iOS player for Z-machine and Glulx interactive fiction. Diction reads the game aloud while you play; with the full version you can also give parser commands by voice. A built-in Find browser lists classic games from IFDB that are a click away.

Features

• Narration. The free version reads game output with Apple's Accessibility text-to-speech engine. The full version adds a neural voice (Kokoro) for a more natural read.
• Voice commands. (Full version) The mic listens continuously and knows when you've stopped talking; your speech is normalized to parser commands. Recognition favors words from the recent game log to handle unusual names.
• Spoken control commands. Say "game" followed by stop, repeat, windows, window N, keywords, inputs, or input N to control narration and review history without touching the screen.
• Library. Browse and download classic games from IFDB, or open your own story files.

The paywall covers voice features only: games and system-voice narration are free; a single in-app purchase unlocks neural narration and voice commands. Voice commands are free to try in the bundled game.

Architecture

Diction runs the interpreters in JavaScript inside a headless WKWebView, the same autosave-capable web IF stack Lectrote uses, and exchanges RemGlk JSON with the Swift app:

• ZVM (ifvms.js) interprets Z-machine story files.
• Quixe (erkyrath/quixe) interprets Glulx story files, and supplies the classic Glk stack (glkapi, dialog, gi_blorb, gi_dispa, gi_load).
• GlkOte + jQuery render the text, grid, and graphics windows inside the web page; that rendered surface is what the app displays.

The Swift side (Diction/Diction/Interpreter/) drives the page through WebInterpreterHost, decodes the RemGlk JSON into presentation snapshots, answers save/restore filerefs internally, and feeds the styled output to the narration and speech-recognition layers.

Setup

The interpreter JavaScript is built from upstream sources rather than vendored in git. After cloning this repository:

npm install                            # pulls the pinned ifvms (ZVM) package
cd Diction/Interpreters
./scripts/fetch_glk_stack.sh           # vendors Quixe + ZVM + GlkOte + jQuery into the app

The script copies the built JS into Diction/Diction/Interpreter/Resources/ and the license texts into Diction/Diction/Resources/Licenses/. Then open Diction/Diction.xcodeproj and build.

Pinned versions

Component	Source	Pin
Quixe + classic Glk stack	erkyrath/quixe	commit d1710c27b0ac293d40bc67cb1b3de38c1ef7fa9f
ZVM (Z-machine)	ifvms	npm 1.1.6

To update, bump QUIXE_COMMIT in scripts/fetch_glk_stack.sh and/or the ifvms version in package.json, re-run npm install and the fetch script, and verify the app still builds. The script re-applies one local patch: GlkOte's MORE paging is disabled so the buffer scrolls continuously for the voice reader.

Credits

Diction is built on these open-source projects.

Interpreters and Glk stack

Project	Role	License	Author
Quixe	Glulx interpreter + classic Glk stack	MIT	Andrew Plotkin
GlkOte	Web display layer — text, grid + graphics windows	MIT	Andrew Plotkin
jQuery	DOM toolkit required by GlkOte	MIT	jQuery Foundation
ifvms.js (ZVM)	Z-machine interpreter	MIT	Dannii Willis

Neural voice

Project	Role	License	Author
FluidAudio	Neural-voice runtime	Apache 2.0	FluidInference
Kokoro	Neural-voice model	Apache 2.0	hexgrad
MisakiSwift	Phonemizer — Swift port of misaki	Apache 2.0	mlalma
misaki	Phonemizer — upstream	Apache 2.0	hexgrad

The in-app Acknowledgements section (Settings → About) is the source of truth for these credits; full license texts ship in the app and live under Diction/Diction/Resources/Licenses/.

License

Diction itself is MIT-licensed — see LICENSE.