Nvidia Launches Cutting-Edge AI Tools for Speech Recognition and Translation Across 25 European Languages

Nvidia Corporation recently unveiled an extensive open-source dataset named Granary, featuring approximately 1 million hours of multilingual audio to enhance speech recognition and translation for 25 European languages. This dataset positions itself among the largest speech corpora available for these languages. In conjunction with Granary, Nvidia launched two advanced AI models: NVIDIA Canary-1b-v2, optimized for transcribing European languages, and NVIDIA Parakeet-tdt-0.6b-v3, designed for real-time transcription capabilities across all languages in the Granary dataset. These tools are expected to empower developers in creating global AI applications, enabling efficient speech functionalities for use cases such as multilingual chatbots, voice customer service, and instant translation services. Nvidia’s innovations signify a significant leap in AI-driven linguistic technologies, enhancing communication and accessibility across diverse platforms. For ongoing updates and insights, consider subscribing to independent journalism sources like Azernews.

Source link