What is voice cloning?
Voice cloning is the use of artificial intelligence to reproduce a specific person’s voice, so that new speech can be generated that sounds like them saying words they never actually recorded. You provide a sample of the voice and a script; the AI outputs audio of that script in the cloned voice.
Voice cloning is a branch of text to speech. Standard TTS speaks with a fixed library of stock voices. Voice cloning adds a step that captures the characteristics of one voice (its pitch, timbre, accent and speaking style) and steers the speech generator to match it.
How does AI voice cloning work?
Most modern systems follow the same conceptual recipe:
- Learning speech in general. A large neural network is trained on many hours of speech from many speakers, so it learns how language sounds and how voices differ from one another.
- Capturing the target voice. When you upload a sample, a part of the system called a speaker encoder turns it into a compact numerical “fingerprint,” often called a speaker embedding, that describes what makes that voice distinctive.
- Generating new speech. The speech model reads your script and uses the fingerprint to shape the output, so the new audio carries the same vocal identity.
Older approaches required fine-tuning, meaning retraining a model on 30 minutes to several hours of one person’s recordings. Today’s zero-shot cloning skips that: the general model has seen enough voices that a few seconds of a new one is enough to imitate it, which is why results arrive in about a minute rather than days.
How is voice cloning different from voice design and voice changing?
| Technique | What you provide | What you get | Typical use |
|---|---|---|---|
| Voice cloning | A short recording of a voice + a script | New speech in that voice | Narrating content in your own voice |
| Voice design | A written description (age, tone, accent, mood) + a script | A brand-new voice that belongs to no real person | Characters, brand voices, games |
| Voice changing (speech-to-speech) | A spoken recording + a target voice sample | The same words and timing, re-spoken in the target voice | Keeping your own delivery with a different voice |
| Standard TTS | A script + a stock voice choice | Speech in a pre-built voice | Everyday voiceovers and read-aloud |
What is voice cloning used for?
Creator narration
YouTubers and course creators narrate scripts in their own voice without re-recording every take.
Localization
The same voice speaks a video in several languages.
Voice preservation
People facing conditions such as ALS can “bank” their voice to keep using it later.
Script fixes
Patch a misspoken word in a podcast without booking another session.
Games and animation
Prototype dialogue with consenting voice actors’ voices.
Brand voice
Keep a consistent, approved spokesperson voice across ads and phone systems.
Is voice cloning legal and ethical?
Cloning your own voice, or a voice whose owner has clearly agreed, is broadly accepted. Cloning someone else without permission is where ethical and legal problems begin. The core principles are simple: get consent, do not deceive, and do not impersonate.
The legal picture in the US is developing quickly. In general terms:
- Right of publicity. Many states protect a person’s name and likeness from commercial use without permission, and courts have long recognized that a distinctive voice can be part of that likeness.
- Tennessee’s ELVIS Act (Ensuring Likeness Voice and Image Security Act, 2024) explicitly added a person’s voice, including AI-simulated versions, to the rights protected under state law, and can create liability for people who make or distribute unauthorized imitations of someone’s voice.
- Robocalls. The FCC has stated that AI-generated voices in robocalls are covered by existing rules on artificial voices, so they generally need prior consent.
- Federal proposals for a nationwide right over digital replicas of voice and likeness have been introduced in Congress, and other states have passed or are considering their own laws.
This is general information, not legal advice. Laws differ by state and country, and Speechma’s AI voice and cloning policy and acceptable use policy prohibit cloning people without permission, impersonation and fraud.
How do I clone my own voice for free?
- Record a clean sampleRead a few natural sentences for 5–30 seconds in a quiet room. One speaker, no music.
- Upload it to SpeechmaOpen voice cloning and upload the file in any common format, or paste a link.
- Write the scriptType up to 1,000 characters, writing numbers out as words.
- Choose language and speedPick from 10 languages (English, Spanish, French, German, Italian, Portuguese, Russian, Chinese, Japanese, Korean) and a speed from 0.8× to 1.25×.
- Generate and downloadEnter the security code, listen, and download. Your sample is not stored on Speechma servers.
Frequently asked questions
How much audio do you need to clone a voice?
Is it legal to clone someone’s voice?
Can I clone a celebrity’s voice?
How can I tell if a voice is cloned?
Can a cloned voice speak another language?
What is the difference between voice cloning and deepfakes?
Make better voiceovers, faster
- New voices and features the day they launch
- Script templates and tricks that make AI voices sound human
- Creator stories, community picks and early access
