How does voice cloning work?
A cloning system is a speech synthesis model given a sample of the target voice. It learns the voice’s pitch, timbre, accent and rhythm from the sample and applies them to new speech. Some systems need hours of studio recording, and others work from a short clip.
Responsible providers ask for proof of consent. OpenAI’s custom voices, as documented on 8 October 2026, are limited to eligible customers and need two recordings: the voice’s owner reading a consent statement, and a reference sample of no more than 30 seconds from the same speaker.
Example: a training company records its presenter reading a consent statement and a half-minute sample, then uses the cloned voice to narrate new course videos.
What are the risks of voice cloning?
The US Federal Trade Commission notes the technology’s promise, including for people who have lost their voices through accident or illness, and its risks: extortion scams aimed at families and small businesses, and voice artists’ voices used in ways that threaten their livelihoods.
In the EU, Article 50 of the AI Act requires providers of AI systems that generate synthetic audio to mark it in a machine-readable way, and deployers of AI that generates audio constituting a deep fake to disclose that it was artificially generated.
For a business, the practical questions are whether the speaker has agreed, whether listeners are told the voice is synthetic and who can use the clone.
Does Kawl Now offer voice cloning?
Kawl Now doesn’t offer voice cloning or custom voices. Each agent speaks in one of two preset voices from OpenAI: Marin, warm and expressive, or Cedar, calm and grounded.
You can’t upload a recording to make a voice, and Kawl Now doesn’t record call audio, so it holds no voice samples from you or your callers. Transcripts are saved as text.
Callers are told before every call that they’re speaking with AI, and the agent is told to say so as well, so callers know the voice is synthetic.
Common questions
Can I use a cloned voice for my voice agent?
Some voice platforms allow it, usually with proof that the speaker has agreed. Kawl Now doesn’t: its agents use one of two preset voices, Marin or Cedar.
How much audio does voice cloning need?
It depends on the provider, from hours of studio recording to a short clip. OpenAI’s custom voices, for example, use a reference sample of 30 seconds or less, plus a separate consent recording from the same person.