How do you make a voice clone with Samsung Galaxy AI?
Go to Settings > Advanced features > Bixby > Language and voice style > Create Custom Voice, install the Bixby Custom Voice Creator from the Galaxy Store, then read the on-screen sentences aloud. It clones only your own voice, recorded live — Samsung ships no tool to clone anyone else's — and the result is used by Bixby Text Call.
Why — the first-principles explanation
The first thing to understand is that "Galaxy AI" has no general-purpose voice cloner. The only voice cloning Samsung ships is a Bixby feature called Custom Voice, and it does exactly one narrow job: it makes Bixby speak in your voice when it answers calls for you via Bixby Text Call. You type a reply, and instead of sending a text, Bixby reads it aloud to the caller in a voice that sounds like you. That's the whole product. If you came here hoping to paste in an audio file of a friend, a singer, or a celebrity, Samsung's answer is no — not a limit you can work around, but the design.
Why is it built that way? Because the enrollment is the consent. You can't upload a file; you have to sit there and read sentences off the screen into the microphone, live, on your own unlocked phone. That single design choice makes it structurally hard to clone someone who isn't present and willing. Cloud voice cloners that accept an uploaded clip have the opposite property — the person being cloned never has to know. Samsung swapped convenience for a consent gate, and the awkward read-ten-sentences-in-a-quiet-room ritual is that gate.
So why is a handful of sentences enough to sound like you? Because you're not teaching the model to speak — it already knows how. Modern text-to-speech models are pre-trained on thousands of hours of many speakers, so they've already learned pronunciation, rhythm, and intonation in the abstract. What they lack is you: your pitch, your timbre, the specific resonance of your throat and mouth. Your recordings supply a compact speaker profile that steers the existing engine. It's closer to picking a paint color than building the wall — which is why a minute of reading gets you most of the way, and why the clone handles words you never recorded.
One fair warning about expectations. This is a Bixby feature, not a headline Galaxy AI feature, and Samsung documents it thinly — its official support pages give the menu path but never state how many sentences you'll read, how long it takes, what battery level you need, or which devices and languages are supported (checked July 2026). Availability varies by region and language, and the custom voice doesn't work everywhere on the phone; it's aimed at Bixby Text Call and some text-to-speech output. If the menu item isn't there, that's usually region or Bixby version, not you doing it wrong.
An example that makes it click
Think of the difference between two ways to get a key to your house. One: a locksmith comes to your door, checks that you live there, and cuts a key while you watch. Two: someone mails a photo of your key to a website and gets a copy in the post.
Samsung's voice cloning is the locksmith. You have to be standing there, phone unlocked, reading sentences out loud — so the only voice you can copy is the one attached to the mouth in the room. The upload-a-clip services are the photo-in-the-mail version, and that's exactly why they're the ones people worry about. Same technology, completely different front door.
How to do it
- Open Bixby's settings: go to Settings > Advanced features > Bixby, or press and hold the Side (Bixby) button and open the settings from there.
- Tap "Language and voice style."
- Tap "Create Custom Voice." If you don't see it, your region, language or Bixby version may not support it.
- Tap "Download" on the pop-up, then tap Install or Update in the Galaxy Store to get the Bixby Custom Voice Creator app.
- When it finishes, tap "Open," allow the microphone permission, review the information screen, and tap "Start."
- Read the sentences shown on screen aloud, in a quiet room, at a normal speaking pace. Hands-on reports put this at roughly ten sentences and a couple of minutes; Samsung's own documentation doesn't state the number.
- Give the voice a name and let Bixby finish processing it.
- Turn it on where it's used: Bixby Text Call will now read your typed replies to callers in your custom voice.
Key facts
- The menu path is Settings > Advanced features > Bixby > Language and voice style > Create Custom Voice, per Samsung's official US support page (checked 2026-07).
- The feature requires downloading a separate app — Bixby Custom Voice Creator — from the Galaxy Store when you first tap Create Custom Voice.
- Enrollment is live-recording only: you read sentences shown on screen into the microphone. There is no option to upload an audio file, so you cannot clone another person's voice.
- The custom voice's main purpose is Bixby Text Call, which reads your typed replies aloud to callers in your voice; Samsung notes it does not work everywhere on the device.
- Samsung's official support pages document the menu path but do not specify the number of sentences, recording time, battery requirement, or supported devices — verified across two Samsung support pages, 2026-07.
- Voice styles and availability vary by region and language; Samsung notes some screens appear in English if your language isn't supported.
▶ The 60-second explainer (script)
How do you make a voice clone with Samsung Galaxy AI? Here's the path: Settings, Advanced features, Bixby, Language and voice style, Create Custom Voice. Tap Download, install the Bixby Custom Voice Creator from the Galaxy Store, open it, allow the microphone, hit Start, and read the sentences it shows you — quiet room, normal pace. Name the voice, and Bixby learns it. But here's the part nobody tells you up front. Galaxy AI has no general voice cloner. This is a Bixby feature, and it does exactly one job: it makes Bixby speak in your voice when it answers calls for you through Bixby Text Call. You type a reply, and the caller hears it read out in a voice that sounds like you. That's the whole product. And notice what you can't do — you cannot upload an audio file. There's no import button. You have to sit there and read, live, on your own unlocked phone. That's not a missing feature. That's the point. The enrollment is the consent. Cloud cloners that take an uploaded clip can copy someone who never agreed and never finds out. Samsung traded convenience for a consent gate, and the awkward reading ritual is the gate. So why is a minute of reading enough to sound like you? Because you're not teaching it to speak — it already knows how. These models are pre-trained on thousands of speakers, so they've learned rhythm and pronunciation in the abstract. What they're missing is your timbre. Your recordings just steer the engine. It's picking a paint color, not building the wall. One honest warning: Samsung documents this thinly. Its support pages give the menu path and not much else — no sentence count, no supported device list. If you don't see the option, it's probably your region or Bixby version.
What authoritative sources say
People also ask
Can I clone someone else's voice with Galaxy AI?
No. Enrollment is live-recording only — you read sentences into the mic on your own unlocked phone, and there's no option to upload an audio file. Samsung ships no tool to clone another person.
What is the custom voice actually for?
Bixby Text Call. When you answer a call by typing, Bixby reads your reply aloud to the caller in your voice instead of sending it as a text. It doesn't work everywhere on the phone.
I don't see "Create Custom Voice" in my settings.
That's usually region, language or Bixby version. Samsung says available settings and language options vary by Bixby version, and voice styles vary by location — it isn't offered everywhere.
How many sentences do I have to read?
Samsung doesn't say. Its official support pages give the menu path but don't state the sentence count, recording time, or battery requirement. Hands-on coverage puts it at about ten sentences.
Can I use this to sing or make music?
No. It's a text-to-speech voice for Bixby, not a singing-voice model. Music voice cloning is a different category of tool entirely.