OpenAI’s new voice synthesizer can copy your voice from just 15 seconds of audio
OpenAI has been rapidly developing its ChatGPT generative AI chatbot and Sora AI video creator over the last year, and it’s now got a new artificial intelligence tool to show off: Voice Generation, which can create synthetic voices from just 15 seconds of audio.
In a blog post (via The Verge), OpenAI says it’s been running “a small-scale preview” of Voice Engine, which has been in development since late 2022. It’s actually already being used in the Read Aloud feature in the ChatGPT app, which (as the name suggests) reads out answers to you.
Once you’ve trained the voice from a 15-second sample, you can then get it to read out any text you like, in an “emotive and realistic” way. OpenAI says it could be used for educational purposes, for translating podcasts into new languages, for reaching remote communities, and for supporting people who are non-verbal.
This isn’t something everyone can use right now, but you can go and listen to the samples created by Voice Engine. The clips OpenAI has published sound pretty impressive, though there is a slight robotic and stilted edge to them.
Safety first
Worries about misuse are the main reason Voice Engine is only in a limited preview for now: OpenAI says it wants to do more research into how it can protect tools like this from being used to spread misinformation and copy voices without consent.
“We hope to start a dialogue on the responsible deployment of synthetic voices, and how society can adapt to these new capabilities,” says OpenAI. “Based on these conversations and the results of these small scale tests, we will make a more informed decision about whether and how to deploy this technology at scale.”
With major elections due in both the US and UK this year, and generative AI tools getting more advanced all the time, it’s a concern across every type of AI content – audio, text, and video – and it’s getting increasingly difficult to know what to trust.
As OpenAI itself points out, this has the potential to cause problems with voice authentication measures, and scams where you might not know who you’re talking to over the phone, or who’s left you a voicemail. These aren’t easy issues to solve – but we’re going to have to find ways to deal with them.
You might also like
OpenAI has been rapidly developing its ChatGPT generative AI chatbot and Sora AI video creator over the last year, and it’s now got a new artificial intelligence tool to show off: Voice Generation, which can create synthetic voices from just 15 seconds of audio. In a blog post (via The…
