Get started
Step by step
- 1
Check whose voice it is
Your own voice, or the voice of someone who has given you explicit written permission to clone it and use it the way you intend. Nothing else.
- 2
Record a clean sample
10 to 30 seconds of one speaker in a quiet room. Record in the browser or upload a file of up to 20 MB.
- 3
Pick the language
Choose the language you speak in the sample. The clone learns how you sound in that language, so record in the language you plan to use most.
- 4
Name it
Your own descriptive name, such as "My narration voice". Names that point to a public figure, celebrity or politician are refused before anything is stored or charged. You can add a photo so the voice is easy to spot in your list.
- 5
Confirm consent
Tick the box that confirms the voice is yours or that you have permission. Without it the clone does not start, and KreatorFlow stores the date and time of that confirmation with the voice.
- 6
Wait for training
The voice shows as Training, then Ready. The list refreshes on its own while it trains, so you can leave the page open and keep working.
- 7
Use it in Text to Speech
When training finishes, the voice appears under "My Voices" and reads any script at 2 credits per 1,000 characters.
A paragraph to read for your sample
Varied sentences with questions, numbers and a pause give the clone more to learn from than one repeated phrase. Read it at your normal pace.
Sample script
Hi, this is my voice, recorded for my own videos. Today I want to show you three quick ideas. First, keep it simple. Second, say it like you would to a friend. And third? Don't be afraid to pause. It's twenty past four on a Tuesday, the coffee is still warm, and we're ready to start.
What is not allowed
The voice cloning rules in the Acceptable Use Policy apply to every cloned voice. KreatorFlow may remove voices and suspend accounts that break them.
- Voices of public figures, celebrities or politicians without their written permission
- Impersonating anyone, or deceiving listeners about who is speaking
- Fraud, scams and fake endorsements
- Political deepfakes, such as invented statements by candidates or officials
- Voices of minors
Cloning someone else's voice: get permission in writing
A spoken "sure, go ahead" is not enough. If the voice is not yours, get written permission before you record anything, and keep it. A short message or email works if it says three things clearly.
If the person later asks you to stop, delete the voice and stop using anything made with it.
- Who gives permission, and that the voice is theirs
- What the voice will be used for, such as narration for your company's training videos
- How long the permission lasts, and that they can withdraw it
Recording setup: what actually makes a difference
You do not need a studio. A phone or laptop microphone in the right room beats an expensive microphone in the wrong one.
A small room with soft things in it (a bedroom with curtains, a wardrobe full of clothes) has far less echo than a kitchen or an empty office. Sit about a hand's width to a forearm's length from the microphone, and switch off fans, air conditioning and notifications before you start.
- Do one test recording and listen back on headphones before the real take
- Keep the same distance from the microphone the whole time
- Drink some water first: dry mouth clicks end up in the clone
- If you stumble, stop and record the whole sample again rather than editing it
Tips for a clean sample
The quality of a clone depends mostly on the sample.
- A quiet room without echo, close to the microphone.
- One speaker only, with no music or background noise.
- Speak naturally at your normal pace.
- Read a varied paragraph rather than repeating one phrase.
- Keep the length between the recommended minimum and maximum.
If the clone does not sound like you
Almost every problem traces back to the sample. Fix the recording, clone again, and compare the two on the same script.
- Sounds hollow or distant: there was echo in the room, or you were too far from the microphone
- Hiss or hum in every line: background noise in the sample, such as a fan or a fridge
- Flat or robotic: the sample was read too carefully. Read it like you are talking to someone
- Wrong accent: the language picked for the clone did not match the language you spoke
- Status shows Failed: the credits are refunded. Record a cleaner sample and try again
Using your cloned voice
A cloned voice works like any preset voice in Text to Speech: the speech rate, the style instruction and the Advanced settings all apply. It is also available in two-speaker dialogues, so you can talk to a preset voice or to a second clone that you have permission for.
Be open with your audience. If a video uses a cloned voice, say so where it matters, for example in the description of an ad or a narrated explainer.
Deleting a cloned voice
In Voice Replication, each voice in your list has a Delete button. Tap it, then tap "Delete forever" to confirm. The voice is removed from your account and can no longer be used to generate speech. Audio you already downloaded stays on your device.
If someone cloned your voice without permission
Report it to support. Cloning without permission breaks the Acceptable Use Policy, and the consent record stored with every voice helps look into the report.



