Skip to content

User Guide

Welcome to the WakeBuilder User Guide! This section covers everything you need to know to create, train, test, and manage your custom wake word models.

Overview

WakeBuilder provides a complete workflow for creating wake word detection models:

flowchart LR
    A[Choose<br/>Wake Word] --> B[Record<br/>Samples]
    B --> C[Configure<br/>Training]
    C --> D[Train<br/>Model]
    D --> E[Test<br/>Model]
    E --> F[Deploy<br/>Model]

In This Section

  • Creating Your First Model

    Step-by-step guide to creating a new wake word model from scratch.

    Create Model

  • Training Workflow

    Understanding the complete training process and what happens at each stage.

    Training Workflow

  • Testing Models

    How to test your trained models with live audio and file uploads.

    Test Models

  • Model Management

    Managing, downloading, and organizing your trained wake word models.

    Manage Models

  • Web Interface Guide

    Complete walkthrough of the WakeBuilder web interface.

    Web Interface

  • Tips for Best Results

    Expert tips and best practices for creating high-quality wake word models.

    Best Practices

Quick Reference

For most users, the default settings work well:

Setting Default Notes
Recordings 3-5 More recordings improve accuracy
Target Positives 5,000 Balanced training set
Negative Ratio 2.0x Good discrimination
Hard Negative Ratio 4.0x Critical for accuracy
Batch Size 32 Works for most hardware
Max Epochs 100 Early stopping handles termination
Learning Rate 0.0001 Stable training

Typical Training Times

Training is a computationally intensive process. Expect it to take time:

Hardware Expected Time
High-end GPU (RTX 3080+) 30 min - 1.5 hours
Mid-range GPU (RTX 3060) 1 - 2 hours
Entry GPU (GTX 1650) 2 - 4 hours
CPU only 4+ hours (not recommended)

GPU Strongly Recommended

Training on CPU is extremely slow due to the complexity of the AST model and data augmentation pipeline.

Getting Help

If you encounter issues:

  1. Check the Troubleshooting section
  2. Review Best Practices for optimization tips
  3. Open an issue on GitHub