Zonos, a new open-source text-to-speech (TTS) model developed by ZyphraAI, has been released, featuring high-fidelity voice cloning and expressive speech capabilities. The model includes two variants, each with 1.6 billion parameters, utilizing transformer and SSM architectures. It supports real-time voice cloning using 5 to 30 seconds of audio input, and allows for adjustments in speed, pitch, audio quality, and emotional tone. Additionally, it offers enhanced speaker matching through the addition of text and audio prefixes. The model is reported to run at approximately double the real-time rate on an RTX 4090 graphics card. Zonos also supports multilingual output, broadening its applicability in various linguistic contexts.