Voice Design by ElevenLabs generates unique synthetic voices from scratch

ElevenLabs has introduced Voice Design, a groundbreaking generative AI tool that creates unique synthetic voices from scratch. Users can specify core characteristics like gender, age, and accent to generate completely original voices that have never been heard before.

The feature works alongside ElevenLabs' existing voice library of 3,000 high-quality voices, allowing users to fill gaps when they need specific voice characteristics. Each generated voice is unique, even when using the same parameter settings, thanks to built-in randomization.

Technical Details

The system currently supports American, British, African, Australian, and Indian accents, with American and British accents producing the most accurate results. Each voice generation requires a minimum of 100 characters and deducts from the user's account balance.

User Experience

Users have reported mixed experiences with the tool. While some praise its ability to create nearly perfect voice clones that capture accents and mannerisms, others note limitations in emotional range and control options. The voice quality can be inconsistent, with some generations speaking fluently while others exhibit typical AI speech patterns.

Applications

The tool serves various use cases:

1. Authors can create unique character voices for audiobooks
2. News publishers can develop distinctive narrator voices
3. Game developers can generate voices for numerous NPCs
4. Content creators can produce custom voiceovers

Current Limitations

Voice Design currently only works with English language content. Users have expressed concern about the character deduction system for voice testing and the limited control over voice characteristics. The tool also faces challenges with accent consistency and emotional expression in generated voices.

The company continues to develop the feature, with plans to introduce more advanced controls and editing capabilities in future updates.