OpenDictate v0.2.0

OpenDictate v0.2.0 — Cross-Platform Synchronization, FastConformer Streaming & Models Hub

OpenDictate v0.2.0 is a major milestone release bringing full feature parity and synchronization across Linux, macOS, and Windows, a new ultra-low latency streaming model, and a redesigned Models management tab.

---

⚡ Highlights & What's New

1. NVIDIA NeMo FastConformer 80ms (Streaming)

• Added the FastConformer CTC (Streaming) model (nemo-streaming-fastconformer-ctc-en-80ms) to the model catalog.
• Ultra-low latency: Features an 80ms chunk emission rate for near-instantaneous live transcription as you speak.
• Lightweight footprint: ~115M parameters (~110MB download / ~125MB disk footprint).
• Full engine support in StreamingRecognizer with automatic CTC vs Transducer architecture detection.

2. Dedicated Models Management Tab

• Extracted model downloading and switching from the bottom of Settings into a dedicated Models tab in the sidebar.
• Features a top-level System Stats Bar tracking:
  • Active recognition engine
  • Total installed model count
  • Total local disk space consumed
  • GPU hardware accelerator status (CUDA, CoreML, or CPU)

3. Full macOS & Windows Synchronization

• Synthetic Input & Voice Actions: Standardized key injection across all operating systems via native platform backends (enigo + CGEvent on macOS, SendInput on Windows, persistent uinput on Linux).
• Launch-at-Login (Autostart):
  • macOS: Native LaunchAgents plist integration (~/Library/LaunchAgents/com.opendictate.app.plist).
  • Windows: Clean registry Run integration (HKCU\Software\Microsoft\Windows\CurrentVersion\Run).
  • Linux: Freedesktop .desktop autostart.
• Cross-Platform App & Tab Controls: Universal voice commands for app switching (switch_to_app), tab management (press_new_tab, press_close_tab, press_next_tab, press_prev_tab), word/line deletion, and web searches.
• Storage Locations: Standardized app data and model storage (~/Library/Application Support/opendictate on macOS, %LOCALAPPDATA%\opendictate on Windows, $XDG_DATA_HOME on Linux).

4. ⌨️ Wayland Virtual Keyboard Persistence & Hotkey Stability

• Upgraded the Linux uinput virtual keyboard to a persistent singleton device (/dev/uinput), eliminating seat binding dropouts under Wayland compositors (GNOME 46/47, Sway, Hyprland).
• Added clipboard settle synchronization ensuring 100% reliable paste operations in text fields and editors.

---

📦 Platform Installers & Downloads

Platform	Recommended Installer	Details
Linux	OpenDictate_0.2.0_amd64.AppImage / .deb	Portable single-file AppImage or Debian package (x86_64)
macOS	OpenDictate_0.2.0_universal.dmg	Universal Apple Silicon (aarch64) & Intel (x86_64) installer
Windows	OpenDictate_0.2.0_x64-setup.exe / .msi	64-bit Windows installer with bundled ONNX Runtime

---

100% Local & Private

• Zero cloud speech API dependencies.
• Zero telemetry or audio transmission.
• All models run locally on your hardware.