AI Voice Clone vs Text Clone: What Each Can Preserve

1What is an AI voice clone?

A voice clone is a text-to-speech model adapted to one person’s voice. It is trained or tuned on audio samples of that person speaking. Once it exists, you can type any sentence and the system speaks it in a voice that sounds like theirs, including words they never said.

Quality depends on the recordings: clean audio, little background noise and a range of speaking styles help. The US Federal Trade Commission has warned that a scammer may need only a short audio clip, for example from a video posted online, plus a voice-cloning program.

A voice clone on its own does not know anything about the person. It reads out whatever text it is given. Some products combine a voice clone with a chatbot that writes the words, which is where personality comes in.

2What is a text personality model?

A text clone, sometimes called a personality model, learns from what a person wrote. Built from chat history, it picks up their greetings, message length, nicknames, emoji, spelling habits, humour and recurring topics. With a memory index, it can also recall real events from those conversations.

Memory Clone is one example. It builds a local index from exported WhatsApp, Messenger, Instagram or Discord chats and your notes, and learns style per relationship, so a clone writes to each contact the way the person wrote to that contact. Read how style learning works for more detail.

3What each one preserves

Neither one captures a whole person. A voice without their words is a recording artist reading a script. Words without a voice can still feel distant. Both are AI simulations, and both can produce things the person never said.

4Consent and fraud risks of voice clones

Voices are used to prove identity, in phone calls to family, banks and employers. That makes a cloned voice more dangerous than cloned writing.

• Family emergency scams. The FTC describes calls where a cloned voice of a relative claims to be in trouble and asks for money by wire transfer, cryptocurrency or gift cards. Its advice: hang up and call the person back on a number you know is theirs.
• Wider fraud. When it launched a Voice Cloning Challenge, the FTC also listed fake calls from company executives and misuse of creative professionals’ voices as harms.
• Robocalls. In February 2024, the US Federal Communications Commission ruled that AI-generated voices count as “artificial” voices under the Telephone Consumer Protection Act. That makes voice-cloned robocalls without prior consent illegal in the US.
• Consent. A voice is part of someone’s identity. Cloning a living person needs their clear permission. For someone who has died, think about what they would have wanted and how family feel.

Laws on voice and likeness rights differ by country and region, and this is general information, not legal advice. For the wider consent picture, see Is it legal and ethical to make an AI clone of someone? and our Acceptable Use Policy.

5Why Memory Clone is text-based today

Memory Clone currently works only with text: exported chats and notes. It does not use voice notes, photos or video, and it does not clone voices. On our public roadmap, support for audio such as voice notes, photos and video is listed as Planned, while voice cloning and a 3D avatar are listed as Research. Neither is available, and we make no promise about dates.

We started with text for three reasons:

1. Text is where everyday relationships live. Years of messages hold nicknames, in-jokes and shared memories that a voice sample alone cannot provide.
2. Lower misuse risk. A text clone presented as AI is much harder to use for a convincing phone scam.
3. Privacy. Imported originals and the memory index stay on your PC. Cloud processing is off until you switch it on for a clone, and then only a bounded selection of excerpts is sent.

Memory Clone is a Windows app in alpha. You can see current plans on the pricing page.

6How to preserve voice recordings safely now

Whatever you decide about AI, real recordings are precious and easy to lose. Saving them now keeps every option open.

1. Save voice notes from chatsExport WhatsApp chats with media, or save individual voice notes. Our guide on saving WhatsApp voice messages covers iPhone and Android.
2. Collect other audio and videoVoicemails, home videos and recorded calls often contain the clearest speech.
3. Keep the original filesDo not rely on compressed copies. Rename files with the date and a few words, for example “2019-06-14 Dad birthday message”.
4. Store two copies privatelyOne on a computer or external drive, one in trusted cloud storage or with a family member. Turn on device encryption, such as BitLocker on Windows.
5. Avoid posting voices publiclyPublic clips are the easiest material for scammers to clone. Share recordings privately with family instead.
6. Write down the contextA short note about who is speaking and why the recording matters helps future generations.

7Which one is right for you?

• If hearing their voice is what matters most, start by preserving real recordings. A real voicemail is often more comforting than any synthetic voice.
• If you miss their way of talking with you, their jokes and the small daily exchanges, a text clone built from your shared chats may fit better.
• If you are thinking about your own legacy, consider recording stories in your own voice and also keeping your messages. A clone of yourself is the clearest case for consent.

For grief-related decisions, go gently. Our guide on talking to a loved one who died with AI can help, and our wellbeing guidance lists free helplines.