We have definitely and very officially reached the tipping point on AI-generated voices. They are very, very good. And ElevenLabs is leading the way.
So if you're looking for some narration work, a video voiceover or something like that - give ElevenLabs a shot.
True human-quality voices
ElevenLabs offers AI-generated voices through VoiceLab, where you can pick out a few attributes that design your voice. Currently, they offer these attributes:
- Gender (male, female)
- Age (young, middle aged, old)
- Accent (American, British, African, Australian, Indian)
- Accent strength (spectrum from low to high)
Then, you type in some text as a sample and hit generate.
And if you don't like that particular variation, hit generate again. And again, and again. It's unlimited voices even if you have the exact same settings.
ElevenLabs' output was the first time we stopped and said, "Ok, wow. They did it." While some generations will have some slight issues - timing, tone inflection, etc. - the vast majority of the created voices are excellent and feel so real.
See an example with this podcast episode.
Clone your voice in 1 minute
ElevenLabs also supports voice cloning. Upload just 1 minute of audio without background noise, and ElevenLabs produces a voice model that can then be used to read out a custom script. Other tools require hours of audio.
It's pretty phenomenal. While you may not need it for making slight edits to a podcast (see Descript's Overdub), you could totally replace any voiceover work you're doing. Recording internal training videos or walkthroughs might be a useful starting point there.
Accessibly priced
ElevenLabs lets you try all the fun features without having to pay. If you decide you need more, your usage scales up to ~40 hours of generated audio per month and 660 custom voices - those come with the $330/month plan.
We're huge fans of ElevenLabs. Go check them out for any voice work you need done.
