OmniVoice Brings Free Voice Cloning to 600+ Languages

An open-source text-to-speech model developed by the k2-fsa team and Xiaomi now lets users clone voices across 646 languages without cost or training, using only a 3–25 second audio sample. Released on March 31, 2026, OmniVoice represents a major shift toward democratizing voice synthesis—previously a capability restricted to well-funded labs and commercial platforms.