NVIDIA Corporation Releases Speech Recognition AI for 40 Languages as Open Weights [Global AI Briefing]

[Financial News] NVIDIA Corporation has unveiled a speech recognition artificial intelligence (AI) model that converts recordings in 40 languages into text and also provides translation. Its strategy is to lower barriers to entry in the speech AI market by making the model available as open weights, so developers can download it and use it in their own services, while also expanding its AI development ecosystem.
The model can also be used in South Korea to develop multilingual AI speech recognition services, including Korean. It is expected to be useful in developing enterprise AI services for tasks such as automatically generating minutes for global meetings, analyzing customer service records involving foreign consumers, and supporting multilingual work.

NVIDIA Corporation officially announced on the 28th (local time) that it had released its speech recognition AI model, “Nemotron 3.5 Transcribe.” The model converts recorded speech into text and also supports translation among 31 languages, with English as the reference language.
It can distinguish up to eight speakers in a single recording, making it useful for tasks such as preparing meeting minutes and analyzing customer service records.
A key feature is the ability to configure the model to recognize speech with reference to lists of technical terms or proper nouns. This allows companies to develop speech recognition services tailored to specific fields such as healthcare, finance, and law.
However, Transcribe analyzes recorded files rather than processing live calls or broadcast audio in real time.
NVIDIA Corporation has made Transcribe available as open weights so developers can download the model weights, run them in their own environments, and adapt the model for specific tasks as needed.
This reflects the spread of NVIDIA Corporation’s strategy to use open-weight models to increase its influence not only in AI chips but also in the AI development ecosystem.
[email protected] Lee Gu-sun Reporter