alfred-mobile/VOSK_MODEL_SETUP.md

Vosk Model Setup Instructions

Step 1: Download the Model

Download the small English model from Vosk:

Direct Link: https://alphacephei.com/vosk/models/vosk-model-small-en-us-0.15.zip

Size: ~40 MB

Step 2: Extract the Model

1. Extract vosk-model-small-en-us-0.15.zip
2. You should have a folder named vosk-model-small-en-us-0.15

Step 3: Add to Android Project

1. Create the assets folder if it doesn't exist:

   mkdir -p ~/.openclaw/workspace/alfred-mobile/app/src/main/assets
   

2. Move the extracted model folder:

   mv ~/Downloads/vosk-model-small-en-us-0.15 ~/.openclaw/workspace/alfred-mobile/app/src/main/assets/
   

3. Verify the structure:

   app/src/main/assets/
   └── vosk-model-small-en-us-0.15/
       ├── am/
       ├── conf/
       ├── graph/
       ├── ivector/
       └── README
   

Step 4: Rebuild the App

The model will be bundled with the APK. This increases the app size by ~40 MB but allows completely offline wake word detection.

Alternative: Smaller Model

If 40 MB is too large, you can use an even smaller model:

vosk-model-small-en-us-0.4 (~10 MB)

• Link: https://alphacephei.com/vosk/models/vosk-model-small-en-us-0.4.zip
• Less accurate but much smaller
• Update the folder name in WakeWordManager.kt to match

Verification

Once the model is in place, the app will:

1. Automatically unpack it to internal storage on first run
2. Load it into memory
3. Start listening for "alfred", "hey alfred", or "ok alfred"

---

Ready to download the model? Let me know when it's in place and we'll continue with the UI integration!