Nvidia Launches Open-Source Transcription AI Model Supporting 40 Languages

28 Oct 2026, 8:30 pm · 2h ago · 1 min read · TechCrunch AINew

Nvidia has launched a free, open-source speech-to-text artificial intelligence model capable of transcribing audio across up to 40 languages. The model is specifically designed for processing pre-recorded audio files rather than live streams, making it immediately viable for meeting notes and medical transcription software workflows. While it expands open access to high-accuracy multilingual audio transcription for developers and enterprise software builders, the lack of real-time streaming capabilities leaves live voice generation and low-latency interaction use cases open to specialized commercial competitors such as ElevenLabs and OpenAI.