Nvidia has released an open AI model, Nemotron 3.5 Transcription, to transcribe speech in 40 languages, including Russian, Kommersant writes. The system is able to process audio recordings up to an hour long, identify different participants in the conversation and translate text. According to the developer, the model can work without an internet connection, and the published weights allow businesses to adapt it to create their own translation tools and process support requests.