Murmur

Privacy-first BYOK (Bring Your Own Key) voice typing application built with Tauri 2 and Svelte 5.

• Cloud providers: ElevenLabs, OpenAI Whisper, Groq — bring your own API key
• On-device: Apple Speech recognition (macOS, no API key needed)
• Self-hosted: any OpenAI-compatible Whisper API (whisper.cpp, faster-whisper, LocalAI)

• Cloud APIs: OpenAI, Claude, Gemini, or custom OpenAI-compatible endpoints
• On-device: Apple Foundation Models — no API key needed
• CLI tools: gemini-cli, copilot-cli for local processing
• Voice commands: shorten, translate, change tone, generate replies
• Personal dictionary for custom terms and aliases

• Floating glassmorphism overlay with waveform visualization
• System tray with configurable global hotkey (default `Ctrl+``)
• Transcription history with search and persistent storage
• Output to clipboard, keyboard simulation, or both
• Audio cues for recording start/stop and errors

• BYOK — all data goes directly to your chosen providers
• On-device alternatives for both STT and LLM require no cloud at all
• Auto-updater for seamless in-app updates

murmur/
├── Cargo.toml                    # Workspace root
├── crates/
│   ├── lt-core/                  # Domain types + traits
│   │   └── src/
│   │       ├── stt.rs            # SttProvider trait, TranscriptionEvent, AudioChunk
│   │       ├── llm.rs            # LlmProcessor trait, ProcessingTask, ProcessingOutput
│   │       ├── config.rs         # AppConfig (TOML-based)
│   │       ├── dictionary.rs     # PersonalDictionary, DictionaryEntry
│   │       ├── history.rs         # TranscriptionHistory, HistoryEntry
│   │       ├── output.rs         # OutputSink trait, OutputMode
│   │       └── error.rs          # MurmurError
│   ├── lt-audio/                 # Audio capture (cpal + resampling + VAD)
│   ├── lt-stt/                   # STT providers (ElevenLabs, OpenAI, Groq, Custom)
│   ├── lt-stt-apple/             # Swift FFI bridge for Apple SpeechTranscriber
│   ├── lt-llm/                   # LLM post-processing (HTTP APIs + CLI)
│   ├── lt-llm-apple/             # Apple Foundation Models (on-device LLM via Swift FFI)
│   ├── lt-output/                # Output (clipboard + keyboard simulation)
│   ├── lt-pipeline/              # Pipeline orchestration + voice commands
│   └── lt-tauri/                 # Tauri app (IPC, state, events, window)
├── ui/                           # Svelte 5 + TypeScript frontend
│   └── src/
│       ├── components/
│       │   ├── history/          # HistoryPanel (transcription history with search)
│       │   ├── overlay/          # FloatingOverlay, WaveformIndicator, TranscriptionView
│       │   └── settings/         # SettingsPanel (standalone 720x560 window)
│       └── lib/                  # Tauri IPC wrapper
├── config/default.toml           # Default settings template
└── prompts/                      # LLM prompt templates

brew tap hydai/murmur
brew install --cask murmur

After installation, clear the quarantine attribute (unsigned app):

xattr -cr /Applications/Murmur.app

1. Download the .dmg from Releases
2. Drag Murmur.app to Applications
3. Clear the quarantine attribute (unsigned app):

   xattr -cr /Applications/Murmur.app

1. Grant microphone and accessibility permissions when prompted
2. Configure your providers in Settings (system tray → Settings)

• Rust 1.92+
• Node.js 22+
• npm 11+

cargo tauri dev              # Development mode
cargo test --workspace       # Run all tests (~120)
cargo build -p lt-tauri --release  # Release binary
cargo tauri build            # Production bundle (.dmg)

User config: ~/Library/Application Support/com.hydai.Murmur/config.toml

Default template: config/default.toml

• Not on Mac App Store — uses macos-private-api for transparent windows (direct distribution only)
• BYOK — cloud providers require your own API keys; on-device and self-hosted options need none
• Apple STT — requires macOS 26+ for on-device speech recognition models

Licensed under the Apache License, Version 2.0. See LICENSE for details.

See CONTRIBUTING.md for development setup, coding conventions, and how to submit changes.