Orpheus TTS

Orpheus TTS is an open-source, Llama-3b-backed Speech-LLM delivering human-like, expressive audio with zero-shot voice cloning and ultra-low latency.

Developed by Canopy Labs, Orpheus TTS is a state-of-the-art, open-source Speech-LLM built on the Llama-3b backbone: it treats speech as a generative LLM task. The system achieves human-level speech quality, featuring zero-shot voice cloning and guided emotional control via simple tags (e.g., <laugh>, <sigh>). Optimized for real-time applications, it boasts ultra-low streaming latency, which can be reduced to 25–50ms with input streaming, making it ideal for conversational AI, virtual assistants, and dynamic content creation using its available voices like 'tara' and 'leo'.

https://github.com/canopyai/Orpheus-TTS