Nvidia releases new open speech-to-text transcription model
Nvidia released a free speech-to-text model that can transcribe up to 40 languages. According to TechCrunch, the model works only on recorded audio, not live audio streams.
Why it matters
The model is free and supports up to 40 languages, which may lower costs for builders of meeting transcription and medical transcription tools. Its lack of live-audio support means real-time features still require other providers such as ElevenLabs.
Who should care
Meeting transcription developers, medical transcription software builders, multilingual app developers, voice AI product teams
What you can do
Test the model on a sample set of your own recordings in your target languages to measure accuracy before deciding whether to adopt it for batch transcription.