Nvidia has released Nemotron 3.5 Transcribe, a free, open-weight model that converts recorded speech into text in up to 40 languages. The company also released version 2 of Granary, its open-source speech dataset, giving developers more material to build and tune voice applications.
According to Nvidia, the model can translate 31 languages to and from English, handle recordings up to an hour long, and distinguish up to eight speakers in a single recording. The company says it outperforms rivals on benchmarks, including Hugging Face’s Open ASR Leaderboard, though those are Nvidia’s own claims and have yet to be tested independently in this release.
The tools are aimed squarely at developers building meeting and medical transcription products, two areas where speech recognition is already a daily workplace habit. Using the model alongside Granary and Nvidia’s NeMo Speech tools, developers can adapt it to recognize industry-specific vocabulary in each language. One clear limit: Transcribe works with recorded audio, not live streams.
At 30 billion parameters, the model is designed to run on a single chip, from the Jetson AGX Thor robotics controller and the DGX Spark chip behind new laptops up to Blackwell data-center hardware. It joins Nvidia’s growing open-weight voice lineup, which includes the much smaller Parakeet model at 1.1 billion parameters.
The launch lands in a busy month for voice AI. Mistral shipped Voxtral 3 earlier in October, ElevenLabs launched its Realtime model for instant speech in voice agents, and Salesforce released Launa-Omni. For anyone comparing established transcription products, our guides to Otter.ai, Descript, and ElevenLabs show how dedicated tools package similar technology for end users, while our Otter.ai vs Fireflies comparison breaks down two of the most popular meeting options side by side.
Source: TechCrunch
