AI voice pioneer ElevenLabs has opened up access to its highly realistic voice cloning technology called Professional Voice Cloning (PVC). Previously only available to enterprise clients, the new release allows anyone to create a cloned version of their own voice.
This launch comes on the heels of the company's unveiling an expanded range of AI-generated voices for its platform just last week. However, PVC represents a more customized approach, training models on users' own voice samples.
ElevenLabs promises PVC can produce a "perfect digital copy" that is "virtually indistinguishable from the original." To create a custom voice clone, users need to upload at least 30 minutes of high-quality speech samples, with 3 hours being optimal. The company then trains a dedicated AI model on this data to replicate the user's vocal patterns, tone, and inflections.
ElevenLabs highlights PVC's potential for audiobook narration, delivering consistent digital presentations, and providing customized voices for IVR systems. The technology could also empower content creators and podcasters to maintain production schedules even when unable to record.
PVC can be integrated with all of ElevenLabs' existing models, including Eleven Multilingual v1. This means users can create flawless copies of their voices in any of the supported languages, further enhancing the technology's versatility.
Importantly, the company has implemented robust security measures to ensure users can only clone their own voice. This includes a nifty, time-locked voice verification system wherein users must read a text captcha aloud, and the voice profile from this recording is matched with the uploaded voice data.
PVC represents a major step in making high-fidelity voice cloning accessible. As the technology proliferates, it may transform how we represent ourselves and interact in digital spaces. But it also raises concerns about potential misuse that will require ethical foresight from creators like ElevenLabs.