The engine supports multiple languages and voices, configurable speaking rates, and custom voice training through model fine-tuning. It accepts plain text or phoneme input and outputs 16-bit PCM WAV audio. Users can switch voices or load new models dynamically at runtime.