AI voice cloning from a 5–30 second recording
Voice cloning uses a recording as a vocal reference when reading your new script. Choose Voice cloning in the studio, then upload or record your own voice, or a voice you have explicit permission to use, to create new English or Chinese narration.
Open toolUpdated
- What you provide
- An authorized, clear 5–30 second voice sample, a new script and optional delivery instructions.
- What you get
- Private WAV speech guided by the sample, with playback, download and deletion. An exact match to the original voice is not guaranteed.
- Sample length
- 5–30 seconds
- Upload formats
- MP3 / WAV, up to 7 MiB
- New script
- Up to 2,000 characters, primarily English or Chinese
- Result
- WAV · kept for 7 days · currently free for a limited time
What you can make
Your own course narration
Use a clear recording of yourself as reference for new lesson text, then check every sentence.
An authorized brand voice
With explicit permission, explore a similar delivery style for different scripts.
Additional film narration
Generate a candidate for a short extra passage, then adjust volume and timing in your editor.
How to use it
- Prepare your own voice or one you have explicit permission to use. Record 5–30 seconds of clear, natural speech, avoiding music, echo and other speakers.
- Open the studio and choose Voice cloning. Upload MP3/WAV up to 7 MiB, or record on the page. Browser microphone permission is needed for recording.
- Listen to the sample and confirm that it is usable, then confirm voice-use authorization. Enter the new script and, if needed, separate pace and emotion instructions.
- Generate, then listen against the new script. The sample’s delivery can affect the output; assess similarity, pronunciation and rhythm yourself.
- Download the WAV. The server removes the reference sample when the task ends. The browser retains an account-scoped draft, clears it on sign-out and does not restore drafts older than 7 days.
Start with an example
Record a welcome message in a voice based on your own
First record 10–15 seconds of yourself speaking naturally. Use the new welcome text as the script, not as a substitute for the audio sample.
- New script
Welcome to today’s creative session. We will start with a simple idea and, step by step, turn it into something we can see, hear and share.
- Style instructions
Keep the reference voice’s natural character. Use a steady pace, as if introducing a course to a friend. Avoid an exaggerated upward inflection at sentence endings.
Check the result: Check the words and pronunciation before judging vocal suitability. A generated file is not a recording of the person actually saying those words.
Open toolCommon questions
Do I have to upload a recording?
A reference recording is required. Upload MP3/WAV or record in the studio; the resulting sample must meet the 5–30 second and size limits.
Can I save a custom voice permanently?
The studio uses a reference per task and does not promise persistent personal voice-model storage or management. Keep an authorized original sample for later use.
How long are the sample and output kept?
The server removes the sample after completion, failure or cancellation. Generated audio is kept for 7 days. Browser drafts are cleared on sign-out and are not restored after 7 days.