AI Voice Cloning with CosyVoice
Clone voices you own or have permission to use from just a few seconds of audio. CosyVoice is an open-source, zero-shot voice cloning model that reproduces timbre, accent and prosody — then speaks your text in multiple languages.
Enter your text
Limit 120 characters per generation. 3 trial samples · 20s wait · Pro skips wait
Select a voice
Awnie · Kids Storyteller
Warm, maternal and soothing delivery for children’s stories and bedtime reading.
Luna · Conversational
Unleash warm, natural, and expressive storytelling voice.
Athena · Audiobook
Clear, formal, and perfectly cadenced British professional voice.
Angus · Warm Narrator
Warm, rich, and highly conversational male voice, perfect for narrating stories and books.
Seán · Podcast Host
Charismatic and effortless male voice with a friendly Irish lilt, ideal for hosting podcasts and discussions.
Orpheus · Explainer
Clear, confident voiceover for explainer videos, product demos, and YouTube tutorials.
Arcas · Commercial
Persuasive, polished reads for ads, promos, and brand commercials.
Why clone voices with CosyVoice
Zero-shot cloning
Reproduce a target voice from a short reference clip. No fine-tuning or hours of training data required.
Open source & free
CosyVoice is released under Apache-2.0. Run it locally, self-host, or try it free online — no vendor lock-in.
Cross-lingual voices
Clone a voice in one language and have it speak another, keeping the same identity across Chinese, English, Japanese, Korean and more.
Natural prosody
Supervised speech tokens capture rhythm, stress and emotion, so cloned voices sound human, not robotic.
What you can build
Audiobooks & narration
Narrate long-form content in a consistent, recognizable voice.
AI agents & assistants
Give virtual agents a branded, real-time voice with sub-second latency.
Video dubbing
Dub videos into new languages while preserving the original speaker’s identity.
Accessibility
Restore or personalize voices for assistive reading and communication tools.
Voice cloning FAQ
What is zero-shot voice cloning?
Zero-shot voice cloning generates speech in a target voice using only a short reference sample — no per-speaker training. CosyVoice extracts the voice identity on the fly and applies it to any text you provide.
Is CosyVoice voice cloning free and open source?
Yes. CosyVoice is released under the Apache-2.0 license, so you can use, modify and self-host it for free. In-browser voice cloning in the playground above is coming soon.
How much audio do I need to clone a voice?
A few seconds of clean reference audio is usually enough for CosyVoice to capture timbre and speaking style. Longer, higher-quality samples improve similarity.
Can I clone a voice in one language and speak another?
Yes. CosyVoice supports cross-lingual cloning, so a voice recorded in English can speak Chinese, Japanese, Korean and other supported languages while keeping its identity.
Is voice cloning ethical and legal?
Only clone voices you own or have explicit permission to use. Cloning someone’s voice without consent may violate privacy and publicity rights. Use CosyVoice responsibly.