Edge0 has announced the release of Audio8-TTS-Realistic, a cutting-edge voice cloning text-to-speech (TTS) model featuring 0.6 billion parameters. This open-weight model operates under the Apache 2.0 license and is designed for zero-shot voice cloning, allowing it to generate realistic audio without prior training on specific voices.
Supporting 23 languages, Audio8-TTS-Realistic aims to broaden accessibility in voice synthesis applications. A notable enhancement in this model is the inclusion of a final Generative Realism Perceptual Optimization (GRPO) stage, which is driven by a learned realism critic, ensuring that the generated audio closely mimics human-like speech patterns.
This release highlights Edge0’s commitment to advancing TTS technology and providing developers with powerful tools for creating lifelike voice applications. As demand for high-quality voice synthesis continues to grow across various industries, this model positions Edge0 as a significant player in the voice cloning landscape.
Compiled automatically by the Tech AI Newsdesk from public AI-news sources and summarised in our own words.
