macroroni 1.1.3

A speech-to-intent controller for a variety of VUI fields.

macroroni

macroroni is an open-source, speech-to-intent framework for Voice User Interfaces (VUI). It captures real-time microphone audio locally and seamlessly translates spoken commands into structured application states or hardware macros, making it perfect for custom automation or gaming wrappers like OpenAI Gym.

---

Key Features

• Dual Engine Architecture:
  • SingleVoiceController: Uses TF-IDF and Logistic Regression to match speech to a single global intent.
  • SequentialVoiceController: Powered by a fine-tuned BERT Token Classification model to extract strings of directions and actions sequentially.
• Hardware-Accelerated Transcription: Local, high-performance ASR backends:
  • Apple Silicon (MPS): Powered by mlx-whisper.
  • NVIDIA GPU (CUDA): Powered by faster-whisper.
• Asynchronous Listening: Non-blocking background threads combined with ambient noise microphone calibration to ensure minimal execution latency.

---

Installation & Hardware Requirements

macroroni requires Python 3.8+. Install the core package along with your machine's specific hardware backend:

pip install macroroni