Dashie supports voice control through Home Assistant's Assist pipeline, using "Hey Dashie" as the wake word. This guide walks you through setting up Whisper for speech recognition and configuring Dashie to work with your Home Assistant voice assistant.
Voice Control Upgrade
Voice control with "Hey Dashie" wake word detection is available as an upgrade for $14.99. The base Dashie app is free and includes all dashboard features.
Prerequisites
- Home Assistant 2023.5 or later with Assist pipeline
- Dashie installed on your Android device
- Whisper add-on or external Whisper server for speech-to-text
- Microphone access on your Android device
How It Works
Voice commands flow through this pipeline:
Wake Word
Records Audio
Speech-to-Text
Process Command
Dashie handles wake word detection locally on the device. Audio is sent to Home Assistant only after the wake word is detected.
Part 1: Install Whisper in Home Assistant
Whisper is OpenAI's speech recognition model that converts spoken audio to text. Home Assistant offers it as an add-on for easy installation.
Open Add-ons
In Home Assistant, go to Settings → Add-ons → Add-on Store.
Install Whisper
Search for "Whisper" and click on the Whisper add-on. Click Install and wait for it to complete.
Configure Whisper
After installation, go to the add-on's Configuration tab. Set the model based on your hardware:
tiny- Fastest, least accurate (good for low-power devices)base- Good balance of speed and accuracysmall- Better accuracy, slowermedium- High accuracy (requires good hardware)
For most users, base or small works well.
Start Whisper
Go to the Info tab and click Start. Enable "Start on boot" if you want Whisper to run automatically.
Performance Note
Whisper can be CPU-intensive. On a Raspberry Pi 4, the tiny or base model is recommended.
For faster processing, consider running Whisper on a more powerful machine or using GPU acceleration.
Part 2: Set Up Wyoming Protocol
Wyoming is the protocol that connects speech services (like Whisper) to Home Assistant's Assist pipeline. The Whisper add-on includes Wyoming support, but you need to verify it's properly configured.
Verify Wyoming Integration
Go to Settings → Devices & Services → Integrations. Look for a Wyoming Protocol integration. If Whisper is running correctly, it should have been auto-discovered.
Add Wyoming Integration (if not auto-discovered)
If Wyoming isn't listed, click + Add Integration and search for "Wyoming Protocol".
Enter the host as core-whisper (the add-on's hostname) and port 10300.
Confirm Whisper Service
Click on the Wyoming integration to see connected services. You should see Whisper listed as a speech-to-text provider. This confirms Wyoming is bridging Whisper to Home Assistant.
What is Wyoming?
Wyoming is Home Assistant's protocol for voice services. It acts as a bridge between speech-to-text engines (like Whisper), wake word detectors, and text-to-speech services. Without Wyoming, Home Assistant can't communicate with these services.
Part 3: Configure the Assist Pipeline
Now that Wyoming is connecting Whisper to Home Assistant, we need to configure the Assist pipeline to use it. We'll also skip text-to-speech (TTS) since Dashie displays responses on screen.
Open Voice Assistants
Go to Settings → Voice assistants.
Create or Edit an Assistant
Click Add assistant to create a new pipeline, or edit an existing one. Give it a name like "Dashie Voice".
Configure Speech-to-Text
Under Speech-to-text, select Whisper (or "faster-whisper" if you installed that variant).
Set Text-to-Speech to None
Under Text-to-speech, select None. Dashie displays responses visually, so TTS is not needed and would slow down the response.
Save the Pipeline
Click Create or Update to save your voice assistant configuration.
Important: Set TTS to "None"
Setting Text-to-speech to "None" is crucial for Dashie. Since responses are displayed on screen, TTS would only add unnecessary delay and audio output. Dashie handles response display natively.
Part 4: Configure Dashie
Open Dashie Settings
In Dashie, tap the gear icon in the sidebar to open Settings.
Navigate to Voice Settings
Find and tap on Voice Control or Voice Settings.
Enable Voice Control
Toggle on Enable Voice Control. If prompted, grant microphone permissions to the app. A restart might be required.
Select Your Pipeline
Under Assist Pipeline, select the voice assistant you configured (e.g., "Dashie Voice").
Test Voice Control
Say "Hey Dashie" followed by a command like "turn on the living room lights". You should see the wake word indicator, then the command being processed.
Example Voice Commands
Once configured, you can use commands like:
- "Hey Dashie, turn on the kitchen lights"
- "Hey Dashie, what's the temperature outside?"
- "Hey Dashie, lock the front door"
- "Hey Dashie, set the thermostat to 72 degrees"
Custom Commands
You can create custom sentences in Home Assistant to expand what commands are understood. Go to Settings → Voice assistants → Sentences to add custom phrases and actions.
Troubleshooting
Wake word not detected
- Ensure microphone permissions are granted to Dashie
- Check that voice control is enabled in Dashie settings
- Speak clearly and at a normal volume
- Make sure the device isn't muted
Commands not recognized
- Verify Whisper is running in Home Assistant
- Check the Assist pipeline is configured correctly
- Try a simpler command to test connectivity
- Check Home Assistant logs for errors
Slow response times
- Try a smaller Whisper model (
tinyorbase) - Ensure TTS is set to "None" in your pipeline
- Check network latency between device and Home Assistant
- Consider running Whisper on more powerful hardware
No response displayed
- Verify the Assist pipeline is selected in Dashie
- Check that Home Assistant URL is correctly configured
- Look for errors in the Dashie logs
Need More Help?
If you're still having trouble, contact us with details about your setup and any error messages you're seeing.