{
  "id": 16914374,
  "title": "Nvidia releases all-in-one adaptable Nemotron 3.5 audio transcription and translation model",
  "url": "https://urgent.news/2026/10/28/nvidia-releases-all-in-one-adaptable-nemotron-3-5-audio-transcription",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-28T15:00:59.000Z",
  "source": {
    "name": "SiliconANGLE",
    "slug": "siliconangle",
    "url": "https://siliconangle.com/2026/10/28/nvidia-releases-all-in-one-adaptable-nemotron-3-5-audio-transcription-and-translation-model/"
  },
  "original_language": "en",
  "account": "Nvidia has launched an open-source artificial intelligence model called Nemotron 3.5 Transcribe, which excels in multilingual audio transcription and translation. This adaptable model is capable of processing up to 100 hours of audio in 40 languages with the ability to identify and label up to eight speakers. The model, which is a 30-billion parameter mixture-of-experts model with 4 billion active parameters, is compact enough to run on small hardware like the Jetson AGX Thor or on a desktop AI supercomputer such as the DGX Spark. For even faster processing, it can operate on a Blackwell data-center GPU with NVFP4 precision and model compression. The model achieves a high accuracy rate of 30.8% error, which can be reduced to 20.8% with careful prompting. Users can instruct the model to transcribe audio, identify speakers, translate content, or respond to queries about the conversation, all through different prompts. Additionally, they can include specific terminology or dialogue to enhance the model's understanding of the content, thereby improving accuracy. Nvidia developed this model by expanding Granary v1 into Granary v2, an extensive open-source multilingual speech AI dataset. This enhanced dataset covers 843,579 hours across 48 languages, significantly outpacing its predecessor. The new dataset generates three transcripts for each recording, each in different styles that can be requested through Transcribe. This data allows developers to adapt transcription formats and fine-tune model outputs for various languages. Overall, the release of Nemotron 3.5 Transcribe demonstrates Nvidia's commitment to making advanced AI technologies accessible and adaptable for real-world applications in transcription and translation.",
  "summary": "Nvidia Corp. today released a leaderboard-winning open artificial intelligence model for multilingual audio transcription and translation, along with an open data pipeline for fine-tuning Nemotron 3.5 Transcribe. Traditional transcription AI models follow a workflow that detects speech, transcribes it, identifies and labels different speakers, then answers questions about the transcription.…",
  "key_points": [
    "Nvidia releases open-source Nemotron 3.5 Transcribe model",
    "Model transcribes 100 hours of audio in 40 languages",
    "Compact design allows operation on small hardware or data-center GPUs"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "The New Stack",
        "title": "Nvidia launches a promptable transcription model",
        "url": "https://urgent.news/2026/10/28/nvidia-launches-a-promptable-transcription-model",
        "published": "2026-10-28T15:00:00.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}