• Meta introduces a new tool for processing rare digital languages

• The system supports dialects and accents through globally diverse audio data

Meta, through its Fundamental AI Research team, has launched an open-source speech recognition system capable of understanding and transcribing over 1,600 languages, including around 500 languages never before supported by AI tools.

The system aims to provide speech recognition technology for lesser-documented languages, with Alexandr Wang, Meta’s AI chief, stating that the technology represents a major step toward universal AI.

Traditional systems typically focus on dominant languages such as English, Mandarin, and Spanish.

Meta hopes to empower researchers and developers to improve speech recognition and translation technologies.

The system relies on a multilingual model trained on diverse datasets from around the world, enabling it to handle dialects, accents, and various speech patterns effectively.

Meta also collaborated with global initiatives to collect authentic voice samples, ensuring accurate and realistic representation of languages.