Release v0.5.6

Release v0.5.6

Release v0.5.5

What's New

Desktop App CI Builds

Tauri desktop apps are now automatically built and released:

• Linux: AppImage and .deb packages
• macOS: Universal .dmg (works on all Macs via Rosetta 2)

Fixed

• Audio device validation: Daemon now validates configured device exists at startup and auto-clears stale devices
• PipeWire defaults: Better handling of PipeWire's actual default audio source
• Tauri CI: Resolved icon format issues and macOS runner deprecation

Changed

• macOS Intel support: Removed dedicated Intel builds from CI (GitHub deprecated Intel runners)
• Intel Macs now use ARM binaries via Rosetta 2 (standard industry practice)

Downloads

Platform	File	Type
Linux (CUDA)	mojovoice-linux-x64-cuda.tar.gz	CLI (Recommended for NVIDIA GPUs)
Linux (CPU)	mojovoice-linux-x64.tar.gz	CLI (Fallback)
Linux	MojoVoice-linux-x64.AppImage	Desktop App
Linux	MojoVoice-linux-x64.deb	Desktop App (Debian/Ubuntu)
macOS	MojoVoice-macos-arm64.dmg	Desktop App (All Macs)
macOS	mojovoice-macos-arm64.tar.gz	CLI (All Macs)

Note: Intel Macs run ARM binaries via Rosetta 2 with minimal performance impact.

Full Changelog: https://github.com/itsdevcoffee/mojovoice/blob/main/CHANGELOG.md

Release v0.5.4

Release v0.5.4

Release v0.5.3

Release v0.5.3

Release v0.5.2

Release v0.5.2

Release v0.5.1

What's New

Added

• Model card visual meters: Replaced text labels (Slower/Fast/Best Quality) with 5-dot visual indicators for Speed and Quality
• Language badges: Color-coded pills for Multilingual (teal) vs English-only (amber) models
• Ghost download button: Transparent button style that fills with color on hover
• Download progress transformation: Download button morphs into progress bar with real-time stats (size, speed, ETA)
• Model Manager UI polish: Enhanced spacing, discoverability, and visual hierarchy across model cards

Fixed

• Active badge overflow: Prevented badge from overflowing in installed model cards
• Audio device enumeration: Improved Linux audio device detection with better PipeWire support
• JSX syntax error: Fixed spacer div and comment syntax in Model Manager
• CUDA CI build: Made CUDA build non-blocking and fixed verification step
• CI linting: Resolved rustfmt and clippy failures

Downloads

Platform	File	Notes
Linux (CUDA)	mojovoice-linux-x64-cuda.tar.gz	Recommended for NVIDIA GPUs
Linux (CPU)	mojovoice-linux-x64.tar.gz	Fallback for non-NVIDIA systems
macOS (Apple Silicon)	mojovoice-macos-arm64.tar.gz	All M-series Macs (M1, M2, M3, M4, M5, etc.)
macOS (Intel)	mojovoice-macos-intel.tar.gz	Older Intel-based Macs

Full Changelog: https://github.com/itsdevcoffee/mojovoice/blob/main/CHANGELOG.md

v0.5.0

Mojo Voice v0.5.0

A major release with daemon management, transcription history, model management UI, and comprehensive benchmarking tools.

✨ Highlights

• Daemon Subcommands - New CLI structure with daemon up, down, restart, status, logs, pid
• Transcription History - Persistent storage with search, filter, copy, and delete
• Model Management UI - Download, cancel, and delete models from Settings
• Audio Device Selection - Choose input device from Settings UI
• Benchmark CLI - mojovoice benchmark with HTML report generation

Added

• Daemon control UI: Start/Stop/Restart buttons in Dashboard
• Real daemon status: UI shows actual GPU (CUDA/Metal/CPU) and model info
• Expanded model registry: 31 Whisper model variants including distil-whisper-v2
• Waybar offline state: Distinct indicator when daemon isn't running
• Mojo-audio FFI: Replaced Candle's buggy mel spectrogram for accurate frame counts
• Dynamic mel bins: Auto-detection of 80 vs 128 mel bins based on model
• Audio test samples: Organized samples with manifest and Harvard sentences
• justfile: Build automation for build, install, daemon management
• Export Diagnostics: DevTools button to save system/daemon info as JSON
• Real system diagnostics: DevTools shows CPU, RAM, GPU info from hardware
• GGUF quantized models: 7 new quantized variants for lower memory usage
• English-only models: Support for .en Whisper variants (tiny.en, base.en, etc.)
• Config commands: --check and --migrate validate all current config fields

Changed

• Settings UI redesign: Cleaner layout with configurable recording timeout
• Model registry: Switched from GGUF to safetensors format for Candle engine
• Config validation: Uses ℹ (info) vs ⚠ (warning) for optional vs migratable fields

Fixed

• Daemon shutdown blocking listener
• Stale state files on startup
• Stale processing indicator in Waybar
• Mojo-audio RFFT frame count
• Daemon restart race condition
• Daemon start when already running
• Settings tab UI freeze
• Model registry download sizes
• Config migrate unnecessary backups
• Export Diagnostics permissions

Refactored

• Simplified main.rs CLI entry point
• Removed dead whisper module
• Simplified audio capture, output, daemon client, and transcription modules
• Cleaned up legacy binary name checks

---

Full Changelog: https://github.com/itsdevcoffee/hyprvoice/blob/main/CHANGELOG.md

Release v0.4.1

v0.4.1 - Premium Settings Panel

- Complete settings editor with validation
- Native file/folder pickers
- Smart daemon restart with safety checks
- Path validation with real-time feedback
- Dropdown styling fixes

Release v0.4.0

v0.4.0 - Tauri UI with glassmorphic design