The quick answer: how to clone your voice
- Record 10–20 seconds of your normal speaking voice in a quiet room, or pick a clean clip you already have.
- Open the Studio, load the model (a one-time download of about 200 MB) and click Clone a voice.
- Record or upload your sample, name the voice, confirm it is yours, and click Create voice. The clone is ready in seconds.
- Type your text and press Generate, then download the result as a WAV file.
That is all it takes. Want every click explained, with screenshots? Read the full step-by-step walkthrough. The guides below make the difference between a clone that sounds “close” and one that sounds like you.
Before you record: a 2-minute checklist
- Room: a small room with soft furnishings (curtains, a bed, a sofa). Avoid kitchens and bathrooms, where hard surfaces echo.
- Noise: turn off fans, air conditioning and notification sounds. Close the window.
- Microphone: a phone or headset about a hand-width (15–20 cm) from your mouth. Keep the distance steady.
- Voice: speak the way you want the clone to sound. A relaxed reading gives a relaxed clone; an energetic reading gives an energetic one.
- Check: listen back with headphones. If you hear hum, hiss or echo, fix it and record again.
What to read: a sample script
You can read anything, but a paragraph with a mix of sounds, a question and a few natural pauses gives the model the most to work with. Read this at your normal pace. It takes about 15 seconds:
The learning path
Follow these guides in order. Each one covers a single step in depth:
- How to record the perfect voice sample: the room, the microphone and the reading style, with clean vs noisy listening examples.
- The complete voice cloning guide: how cloning works and what decides the quality of a clone.
- How to write scripts for AI voices: punctuation, numbers and names that make the clone sound natural.
- Voice cloning vs text-to-speech: when a built-in preset voice is the better choice.
- Troubleshooting: fixes for model loading, microphone permission, noisy clones and slow generation.
Common first-clone problems
| What you hear | Likely cause | Fix |
|---|---|---|
| Hiss or hum behind the voice | Background noise in the sample | Switch off fans and electronics, record again |
| Distant, echoey voice | Room reverb, mic too far away | Smaller, softer room; mic a hand-width away |
| Doesn't sound like you | Clip too short, or more than one voice in it | 10–20 seconds of you alone, speaking continuously |
| Flat or bored delivery | The sample was read flatly | Record with the energy you want to hear |
| Numbers or names read oddly | The text, not the voice | Write them out the way they are spoken |
Remember: clone your own voice, or a voice you have explicit permission to use.