Grand Diomande Research · Full HTML Reader

SpeakFlow 🎙️

Hold **⌥Space** anywhere on macOS → speak → text appears in whatever app is focused. No cloud, no subscription, no latency.

Language as Infrastructure research note experiment writeup candidate score 24 .md

Full Public Reader

SpeakFlow 🎙️

> Wispr Flow competitor — free, offline, private voice-to-text for macOS + iOS

What It Does

Hold ⌥Space anywhere on macOS → speak → text appears in whatever app is focused. No cloud, no subscription, no latency.

Project Status

Wave	Tasks	Status
Wave 0: Foundation	SF-0.1 → SF-0.4	✅ Done — builds, pipeline wired
Wave 1: Mac Injection	SF-1.1 → SF-1.4	⬜ Queued
Wave 2: AI Cleanup	SF-2.1 → SF-2.3	⬜ Queued
Wave 3: iOS Keyboard	SF-3.1 → SF-3.5	⬜ Queued
Wave 4: Ship	SF-4.1 → SF-4.4	⬜ Queued

Building

bash

# Build macOS app (no code signing needed for dev)
cd Desktop/SpeakFlow
xcodebuild -scheme SpeakFlow -destination 'platform=macOS' build \
  CODE_SIGN_IDENTITY="" CODE_SIGNING_REQUIRED=NO CODE_SIGNING_ALLOWED=NO GENERATE_INFOPLIST_FILE=YES

Architecture

SpeakFlow (macOS menu bar)
  └── AppDelegate         — status bar icon, menu
  └── AppCoordinator      — orchestrates the pipeline
        ├── HotKeyManager  — ⌥Space global hotkey (Carbon API + CGEvent tap)
        └── SpeechService  — SFSpeechRecognizer, on-device, offline

SpeakFlowCore (shared framework)
  ├── SpeechService.swift      — SF-0.2: STT engine
  ├── HotKeyManager.swift      — SF-0.3: global hotkey (hold pattern)
  ├── PermissionsManager.swift — accessibility + speech permissions
  └── TranscriptionResult.swift — shared models

SpeakFlowMobile (iOS host)
  └── Minimal app for keyboard extension distribution

SpeakFlowKeyboard (Wave 3)
  └── TBD — iOS custom keyboard extension

Wave 0 Pipeline

⌥Space held → HotKeyManager.onHotKeyDown
           → SpeechService.startRecording()
              └── AVAudioEngine → SFSpeechAudioBufferRecognitionRequest
                  └── SFSpeechRecognizer (on-device, en-US)
                      └── AsyncStream<TranscriptionResult>
                          └── console (Wave 1: will inject into focused app)
⌥Space released → SpeechService.stopRecording() → finalText

Existing Code Merged

`Desktop/Speak/SpeakReader/` — HotKeyManager, PermissionsManager ported
`Desktop/Speak/services/macos_speech_service.py` — API design reference
`Desktop/cross-script-bridge/ios-keyboard/` — KeyboardViewController (Wave 3 source)
`Desktop/voice-corpus/` — 382 voice notes (Wave 2 training data)

Discord Ops Channel

Track progress: [#speakflow-ops](https://discord.com/channels/1470100197941051623/1473426349849972739)

Promotion Decision

Attach run IDs, datasets, metrics, and reproduction commands.

Source Anchor

NKo/macos/SpeakFlow/README.md

Detected Structure

Method · Evaluation · Code Anchors · Architecture