NVIDIA launches an open-weight model for speech-to-text conversion supporting 40 languages.
- This model supports up to 40 languages, marking a significant step in the growing competition within the voice AI technology market.
- Advanced Translation and Speaker Recognition The new model boasts the ability to translate between English and 31 other languages.
- Integration with NeMo Speech Ecosystem Developers can integrate the new model with the Granary dataset and other tools from the NeMo Speech ecosystem.
- Efficient Performance Across Platforms Despite containing 30 billion parameters, Nvidia stated that the Nemotron 3.5 Transcribe can operate using a single computing chip.
Nvidia Launches Free Open-Weight Model for Audio Transcription
Nvidia has announced the launch of the Nemotron 3.5 Transcribe, a free and open-weight model designed to convert audio recordings into written text. This model supports up to 40 languages, marking a significant step in the growing competition within the voice AI technology market.
New Open-Source Audio Dataset
The company also unveiled the second version of the open-source audio dataset Granary, which enables developers to create specialized applications for speech processing and recognition.
Advanced Translation and Speaker Recognition
The new model boasts the ability to translate between English and 31 other languages. It can process audio recordings lasting up to one hour and distinguish between as many as eight speakers within a single recording. Nvidia confirmed that the model achieves advanced performance in automatic speech recognition tests, enhancing its capability to meet the needs of developers and organizations in areas such as meeting transcription, medical conversation documentation, and multi-speaker audio content processing.
Integration with NeMo Speech Ecosystem
Developers can integrate the new model with the Granary dataset and other tools from the NeMo Speech ecosystem. This integration allows for the development of specialized applications capable of recognizing technical terminology across various sectors and languages.
Efficient Performance Across Platforms
Despite containing 30 billion parameters, Nvidia stated that the Nemotron 3.5 Transcribe can operate using a single computing chip. This feature enables its deployment across a wide range of platforms and devices, from embedded systems to data centers.
Growing Competition in Voice AI Technology
The launch of this new model occurs amid rising global competition in voice AI technologies. Several tech companies are entering this field with advanced models for speech recognition and multilingual voice interaction.
To follow the news in Arabic
