What it is
Speech synthesis indistinguishable from a live voice, and cloning a specific voice from a short recording. The trend made voiceover available to everyone — and at the same time produced a wave of voice deepfakes and scam calls.
Where it came from
ElevenLabs pushed the quality into the mainstream (natural delivery plus a clone from a few minutes of audio), alongside open-source voices. Then came voice inside video models and assistants. By 2025 a voice clone is an ordinary feature, not a lab experiment.
Why it took off
- Voiceover without a narrator or a studio: audiobooks, videos, IVR, podcasts — in minutes.
- Your own voice at scale: record once, then "narrate" hundreds of videos and languages.
- Emotional range in the new models removed the robot from synthesis.
How to use it now
- Narration: ElevenLabs (natural) or open-source voices running locally (free).
- Your own clone for content series and dubbing (see the "AI dubbing" card).
- Finishing: cleanup and level matching in Descript or any editor. What you get: even, professional narration without booking a voice actor.
What to watch out for
- Voice deepfakes and fraud are a real threat (scam calls in "your relative's or boss's voice"). Clone only your own voice, or with explicit consent.
- Imitating a public figure's voice without permission is a legal and reputational risk.
- Agree on a spoken code word with your family as protection against fraud calls.