Repository navigation
v2.12.0
Voice Fingerprint
Recognise only your voice — drop neighbours / family / background audio before transcription. Works for both push-to-talk and hands-free. 100 % on-device, ~25 MB model on first enable. Per-device profiles (AirPods and built-in mic train independently).
macOS 14 compat
- Patched
ggml-metalto disableMTLResidencySetDescriptor(macOS-15-only Objective-C class) — the dylib now loads cleanly on macOS 14, where it previously crashed at startup. - Local LLM (Gemma 4 transform) falls back to CPU on macOS < 15 — slower but functional. macOS 15+ keeps the Metal-accelerated path.
Notch alignment
Status pill now sits flush with the bottom of the notch on macOS 14, matching macOS 15 behaviour. Window level raised above the system menu bar (pattern from Ice / VoiceInk / NotchDrop).
Push-to-talk fix
Stopped emitting empty 1-second silence chunks while PTT was held but the user was no longer speaking — the status indicator no longer pulses uselessly mid-press.
Download the DMG, drag SpeechButton.app into /Applications. brew users: brew upgrade --cask speechbutton (after the cask catches up).