Voice and recording
How to record 10 clean seconds to clone your voice
Prepare a quiet place, a single voice, and a natural reading before creating your profile in Vox.
To clone your voice in Vox, the app accepts a recording of up to 10 seconds. Making good use of that time depends more on clarity than on trying to speak quickly. A sample with just one person, steady volume, and little interference provides a more useful starting point than a rushed excerpt. Results can still vary: careful preparation does not guarantee a perfect copy.
Prepare the space before recording
Turn off music and avoid nearby conversations. If a fan or a window lets in noise, change your position or move to another room. You do not need to buy new equipment to get started; first check what the microphone you already have picks up.
Keep a comfortable, consistent distance. Avoid brushing against the device, covering the microphone, or changing its orientation while you read. If the room produces a lot of echo, look for a space with fewer exposed hard surfaces.
Read naturally
Select the language and follow the sample sentence shown in My Voice. Speak in your usual tone and do not force an accent you do not use. Cartesia recommends samples in the language the person speaks and clear audio.
Do not speed up to fill the ten seconds. Do not add music, effects, or a second voice either. If you make a significant mistake or an interruption is audible, record the sample again before cloning it.
Listen and decide before continuing
Review the audio available on the card: can you understand every sentence? Is there a loud thump? Was the ending cut off? Correct the sample if necessary. When you complete the cloning process, Vox takes you to the listening room.
Try a short passage there and check names and pauses against the written text. A good-quality sample does not eliminate every possible TTS error. Keep your expectations realistic and use only your own voice or a voice whose owner has authorized cloning for that purpose.