Skip to main content

Voice Interface

Voice is the primary interface for DotAgents. Hold to speak, release to act. Your agents listen, think, and execute — all triggered by your voice.

Voice Modes​

DotAgents offers several voice interaction modes:

Hold-to-Record (Dictation)​

The default mode for quick voice input:

  1. Hold Ctrl (macOS/Linux) or Ctrl+/ (Windows)
  2. Speak your request
  3. Release to stop recording
  4. Your speech is transcribed and inserted into the active application

This is pure dictation — the AI transcribes your speech and types it wherever your cursor is.

MCP Agent Mode​

Voice input that triggers full agent execution with tools:

  1. Hold Ctrl+Alt to start recording
  2. Speak your request (e.g., "Search GitHub for recent issues in my repo")
  3. Release Ctrl+Alt to process
  4. The agent reasons about your request, executes MCP tools, and responds
  5. Watch real-time progress as each tool is called

Toggle Dictation (Fn)​

Instead of holding a key, toggle dictation on and off:

  1. Press Fn to start recording
  2. Speak freely
  3. Press Fn again to stop and transcribe

Hands-Free Mode (Mobile)​

On the mobile app, hands-free mode uses Voice Activity Detection (VAD):

  1. Toggle the microphone icon in the chat header
  2. The app listens continuously
  3. When you speak, it transcribes automatically
  4. When you stop speaking, it sends the message
  5. Perfect for driving, cooking, or multitasking

Text Input​

For when voice isn't convenient:

  • Ctrl+T (macOS/Linux) or Ctrl+Shift+T (Windows) opens a text input overlay
  • Type your message and press Enter
  • Same agent processing as voice input

Speech-to-Text (STT)​

DotAgents supports multiple STT providers for transcription:

ProviderModelsSpeedQualityOffline
OpenAIWhisperFastExcellentNo
GroqWhisper (accelerated)Very FastExcellentNo
ParakeetONNX modelFastGoodYes

Language Support​

DotAgents supports 30+ languages for speech recognition:

Afrikaans, Arabic, Armenian, Azerbaijani, Belarusian, Bosnian, Bulgarian, Catalan, Chinese, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, Galician, German, Greek, Hebrew, Hindi, Hungarian, Icelandic, Indonesian, Italian, Japanese, Kannada, Kazakh, Korean, Latvian, Lithuanian, Macedonian, Malay, Marathi, Maori, Nepali, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swahili, Swedish, Tagalog, Tamil, Thai, Turkish, Ukrainian, Urdu, Vietnamese, Welsh.

Configure your preferred language in Settings > General.

Configuring STT​

  1. Go to Settings > General
  2. Under "Speech-to-Text", select your provider
  3. Choose the model variant (if applicable)
  4. Set your preferred language
  5. Test with a voice recording

Text-to-Speech (TTS)​

Agent responses can be spoken aloud with multiple TTS providers:

ProviderVoicesQualitySpeed
OpenAI6 voices (Alloy, Echo, Fable, Onyx, Nova, Shimmer)HighFast
GroqOrpheus voicesHighVery Fast
Google GeminiMultiple voicesHighFast
KittenCustom voicesVariableFast
SupertonicCustom voicesVariableFast

TTS Features​

  • Auto-play — Automatically speak agent responses as they arrive
  • Voice selection — Choose from 50+ AI voices across providers
  • Audio player — Play, pause, and replay TTS output
  • Streaming — Audio begins playing before the full response is generated

Configuring TTS​

  1. Go to Settings > General
  2. Under "Text-to-Speech", select your provider
  3. Choose your preferred voice
  4. Toggle auto-play on/off
  5. Adjust volume and speed (provider-dependent)

Voice Flow Architecture​

              ┌──────────────┐
              │  Microphone  │
              └──────┬───────┘
                     │
              ┌──────▼───────┐
              │  Recording   │
              │  (Hold key)  │
              └──────┬───────┘
                     │
              ┌──────▼───────┐
              │  STT Provider│
              │  (Whisper)   │
              └──────┬───────┘
                     │
              ┌──────▼───────┐
              │  Transcribed │
              │  Text        │
              └──────┬───────┘
                     │
         ┌───────────┼───────────┐
         │                       │
    ┌────▼─────┐          ┌──────▼───────┐
    │ Dictation│          │ Agent Mode   │
    │ Mode     │          │ (MCP Tools)  │
    └────┬─────┘          └──────┬───────┘
         │                       │
    ┌────▼─────┐          ┌──────▼───────┐
    │ Insert   │          │ LLM Engine   │
    │ Text     │          │ (Tool Calls) │
    └──────────┘          └──────┬───────┘
                                 │
                          ┌──────▼───────┐
                          │ Response     │
                          ├──────────────┤
                          │ Text Display │
                          │ TTS Playback │
                          │ Text Insert  │
                          └──────────────┘

Emergency Stop​

If an agent is running a long or undesired operation:

Ctrl+Shift+Escape — Immediately stops all active agent sessions.

This is the kill switch. It aborts all in-flight LLM calls, tool executions, and ACP delegations.

Tips​

  • Short commands work best — "Search for React tutorials" is better than a paragraph of instructions
  • Be specific — "Open the file at src/index.ts" beats "open that file"
  • Use agent mode for actions — Hold Ctrl+Alt when you want the agent to do something, not just transcribe
  • Check your mic — Ensure the correct microphone is selected in your system settings
  • Grant permissions — macOS requires accessibility permissions for keyboard monitoring

Next Steps​