AI Voice Cloning — Cast Real Voices with @Audio References
Upload up to three reference clips, mention them as @Audio1–@Audio3 in your prompt, and Seed Audio 1.0 casts those voices in a complete scene — no model training, no waiting, and the timbre stays consistent even when you Extend the story.
Clone a voice freeVoice cloning built for storytelling
Voices in action
Frequently asked questions
Zero-shot means the model clones a voice from a short reference clip without any per-voice training. Upload the clip, reference it in your prompt, and Seed Audio 1.0 reproduces its timbre in newly generated speech right away.
Up to three. Attach as many as three reference audio clips, then address them in the prompt as @Audio1, @Audio2, and @Audio3 to script a multi-voice dialogue in a single scene.
Each uploaded clip gets a slot: @Audio1 through @Audio3. Write them into the scene like character names — for example "@Audio1 reads the evening news while @Audio2 reacts in the background" — and the generated mix casts each cloned voice accordingly.
No — the two casting methods are mutually exclusive in a single generation. Use reference audio (@Audio1–3) when you need specific real voices, or a single character image when you want the model to infer a fitting voice from a face.
Clean, clearly spoken audio of a single person with minimal background noise or music. The closer the reference matches the tone you want in the output (calm, energetic, formal), the more faithful the cloned performance.
Cloning your own voice or a voice you have explicit permission to use is fine. Cloning someone without consent — celebrities, colleagues, anyone — can violate personality and publicity rights and is against our terms. Get written permission for any voice that is not yours.
Yes, provided you have the rights to the reference voice. Generated content on Seedaudio can be downloaded and used commercially — narration, ads, games, video dubbing — as long as the underlying voice is yours to use.
Up to about two minutes per generation. Use Extend to continue longer scripts — the cloned voices keep the same timbre across every continuation, so chapters and episodes stay coherent.
Voice cloning is part of standard generation — it consumes the same credits as any scene. New accounts get free credits on signup, so you can test your first cloned voices at no cost.
Voice cloning applies to speech and dialogue in Seed Audio scenes. Full songs with vocals come from our separate Song Generation tool, which works from your lyrics with its own vocal styles rather than cloned voices.
All generations land in your library. Download the audio files, or share any work through its public /audio/{id} page — handy for sending voice tests to clients before final delivery.