Acknowledgments¶

WakeBuilder builds upon the outstanding work of the open-source community.

Core Technologies¶

Audio Spectrogram Transformer (AST)¶

The foundation of WakeBuilder's few-shot learning capability.

• Author: Yuan Gong, MIT
• Paper: "AST: Audio Spectrogram Transformer"
• Repository: github.com/YuanGongND/ast
• License: MIT

PyTorch¶

The deep learning framework powering all neural network operations.

• Organization: Meta AI
• Repository: github.com/pytorch/pytorch
• License: BSD-style

Hugging Face Transformers¶

Provides the pre-trained AST model and easy model loading.

• Organization: Hugging Face
• Repository: github.com/huggingface/transformers
• License: Apache 2.0

TTS Engines¶

Piper TTS¶

High-quality offline neural text-to-speech.

• Author: Michael Hansen (Rhasspy)
• Repository: github.com/rhasspy/piper
• License: MIT

Edge TTS¶

Access to Microsoft's neural TTS voices.

• Author: rany2
• Repository: github.com/rany2/edge-tts
• License: GPL-3.0

Kokoro TTS¶

High-quality neural TTS with natural prosody.

• Author: hexgrad
• Repository: github.com/hexgrad/kokoro
• License: Apache 2.0

Coqui TTS¶

Multi-speaker TTS models including VCTK.

• Organization: Coqui AI
• Repository: github.com/coqui-ai/TTS
• License: MPL 2.0

Audio Processing¶

librosa¶

Audio and music analysis library.

• Authors: Brian McFee et al.
• Repository: github.com/librosa/librosa
• License: ISC

soundfile¶

Read and write audio files.

• Repository: github.com/bastibe/python-soundfile
• License: BSD-3-Clause

Web Framework¶

FastAPI¶

Modern, fast web framework for building APIs.

• Author: Sebastián Ramírez
• Repository: github.com/tiangolo/fastapi
• License: MIT

Uvicorn¶

Lightning-fast ASGI server.

• Author: Tom Christie
• Repository: github.com/encode/uvicorn
• License: BSD-3-Clause

Model Export¶

ONNX¶

Open Neural Network Exchange format.

• Organization: ONNX Community
• Repository: github.com/onnx/onnx
• License: Apache 2.0

ONNX Runtime¶

Cross-platform inference engine.

• Organization: Microsoft
• Repository: github.com/microsoft/onnxruntime
• License: MIT

Documentation¶

MkDocs Material¶

Beautiful documentation theme.

• Author: Martin Donath
• Repository: github.com/squidfunk/mkdocs-material
• License: MIT

Datasets¶

Speech Commands¶

The dataset used to pre-train the AST model.

• Creator: Google
• Paper: "Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition"

UNAC (Universal Negative Audio Corpus)¶

Negative sample dataset for training.

• Source: Curated from various public domain audio sources

Special Thanks¶

• The Hugging Face team for making state-of-the-art models accessible
• The Rhasspy community for Piper TTS and voice assistant tools
• Everyone who has contributed to the open-source ML ecosystem