01
What makes a useful voice-cloning sample
The uploader accepts MP3 or WAV recordings between 5 and 30 seconds, with a maximum file size of 5 MB. Five seconds can work, but a relaxed 15–25 second take usually captures more of your everyday pacing and pronunciation. Record one speaker in a quiet room, keep the microphone distance steady, and avoid music, echo, aggressive noise removal, shouting, or whispering. Use ordinary speech rather than singing: the singing scene is available only for supported built-in voices, not cloned voices.
02
How to clone your voice online
Create or sign in to your account, then open the clone panel that is already selected on this page. Upload a supported file or record directly in the browser. Listen to the sample, choose a suitable scene, and confirm that the voice belongs to you or that you have explicit permission. Add the text you want spoken, adjust the speed, and choose MP3 or WAV. When generation finishes, review the audio in the player before downloading it. A failed provider request does not leave you with a fake download or a completed task.
03
Small recording choices improve the result
A phone can provide a good reference when the room is quiet. Place it about a handspan from your mouth and slightly to one side so breath and hard consonants do not hit the microphone directly. Read a natural sentence with several sounds, one short pause, and a question instead of a list of numbers or names. Wear headphones for the final check. If you hear another speaker, television, a fan, clipped peaks, or a hollow room echo, record again before changing generation settings; a cleaner source usually matters more than repeated retries.
04
How the reference recording is handled
The provider request runs through the application server, so the speech API credential never reaches your browser. The reference recording is sent for the current voice-cloning request and is omitted from the saved task log, along with the original text. The generated result may appear in your authenticated history for the configured retention window so you can play or download it again. This design reduces unnecessary storage of the source recording, but you should still avoid uploading sensitive conversations or a voice you are not authorized to use.
05
Voice cloning requires real consent
Only clone your own voice or a voice whose speaker has clearly agreed to this use. A public clip, social post, interview, or podcast episode is not permission by itself. Do not use a cloned voice for celebrity imitation, fraud, impersonation, harassment, misinformation, social engineering, or bypassing voice authentication. Tell listeners when synthetic audio could reasonably be mistaken for a real statement, and keep evidence of permission for client or team projects. The voice use policy explains the boundaries and reporting route in more detail.
