AI Voice Cloning: Clone Your Own Voice and Have It Read Text
Chat shortcut: /voiceclone
With AI voice cloning, you read a few minutes of text aloud, upload the recording and get a text of up to 2,500 characters spoken in your own voice. The clone is created only for this run and deleted afterwards.
With Account
Cost: 17,800 tokens per 1,000 characters or part thereof · The uploaded file is deleted after 24 hours; the result stays in “My media”.
This tool can be used with a free account. Sign up for free
Your voice, without speaking into the microphone every time
With voice cloning, a model learns from a short voice sample how your voice sounds: timbre, pitch, speaking rhythm. It then reads your text (up to 2,500 characters per run) in that voice. Unlike text to speech with library voices, this sounds like you, and unlike the Voice Changer, you do not have to record the text yourself.
You need this if you regularly produce audio in your voice but do not want to record every time: corrections in a podcast episode, voiceover for explainer videos, training content or a read-aloud version of your own text. You only change the text; because a new clone is created for each run, similarity and emphasis can vary slightly between runs.
The clone is generated only for that run and then deleted at the provider; there is no permanent voice library. That costs a fresh upload per run, but avoids permanent storage of your voice as a clone.
How it works
- 1
Record and upload a sample
At least 10 seconds, ideally 1 to 3 minutes read calmly and clearly; upload as MP3, M4A, WAV, OGG or FLAC up to 50 MB and 5 minutes.
- 2
Enter the text
Enter the text to be spoken, up to 2,500 characters; punctuation and paragraphs control pauses.
- 3
Remove noise and confirm
Leave "Remove background noise from the sample" switched on and check the box to confirm that it is your voice or that you have permission.
- 4
Listen and save
Start the run, listen to the result and download it as MP3 or carry on editing with MP3 Cutter.
Typical uses
Podcast correction
Replace a misspoken sentence or number afterwards in your own voice, without re-recording.
Voiceover for explainer videos
Write the script, output it in your own voice, and when things change only adjust the text.
Courses and training
Produce lessons in a consistent voice, even if the content comes together over weeks.
Read your own texts aloud
Offer a blog post, newsletter or book chapter as an audio version in your own voice.
Good to know
- Sample: MP3, M4A, WAV, OGG or FLAC up to 50 MB, at least 10 seconds and at most 5 minutes; recommended are 1 to 3 minutes of clear speech from a single speaker without music or reverb.
- The switch "Remove background noise from the sample" is on by default and helps with hiss or room sound.
- Without the consent checkbox, no run starts; only your own voice or one with explicit permission is allowed. Checking it is a declaration, not an identity check.
- The clone is generated only for this run and then deleted at the provider; if deletion fails repeatedly, this is logged. The sample is deleted on our end after 24 hours (hourly cleanup); the MP3 stays in your media library until you remove it there.
- Text up to 2,500 characters per run; cost: a fixed token amount per 1,000 characters or part thereof, cloning costs nothing extra. Account required, no subscription needed.
- Use of the result under the terms of the audio provider; when publishing, observe labeling requirements for AI content.
Frequently asked questions
Whose voice may I clone?
Only your own or that of a person who has explicitly allowed you to; the voice is part of personality rights. By checking the box you declare this; it is not an identity check, and misuse leads to the account being blocked.
How do I record a good sample?
A quiet room, one speaker, no reverb, no music, at least 10 seconds, better 1 to 3 minutes in the tone you want to hear in the result. A consistent microphone distance and varied sentences help more than a long recording. Phone recordings are fine.
Is my voice stored?
Not as a permanent clone: for the run, an instant clone is created at the provider, the text is spoken and the clone is deleted afterwards; if deletion fails, it is retried and otherwise logged. The sample is deleted on our end after 24 hours (hourly cleanup); the MP3 stays in your media library.
How do I control pauses and emphasis?
Through the text: commas, periods and paragraphs create pauses, question marks raise the voice. The clone takes the tone from the sample; emphasis and similarity can vary slightly between runs.
Does the clone speak other languages too?
The model can speak text in other languages with your timbre; we have not checked accent and quality for each language. Best to test with a short text.
What does it cost?
A fixed token amount per 1,000 characters of text or part thereof, shown above the tool; one run takes up to 2,500 characters. Cloning itself costs nothing extra. Account required.
Do I have to label AI voices?
Generated speech counts as synthetic. When publishing, labeling requirements may apply, especially if a real person is imitated (in the EU, for example, under Art. 50 of the AI Act).