NVIDIA Releases Nemotron 3.5 Transcribe for Multilingual Audio Analysis
In short: NVIDIA has introduced Nemotron 3.5 Transcribe, an open-weight, promptable model for speech-to-text, speaker diarization, and translation. It utilizes a hybrid Mamba-Transformer architecture to achieve industry-leading performance with only 4 billion active parameters.
This summary was generated automatically by AI from Hugging Face's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1Added support for transcription in 40 languages and speaker attribution in 13.
- 2Introduced granular prompt-level control over transcription formatting and disfluencies.
- 3Released Granary v2, an open data pipeline for custom model training and adaptation.
- 4Optimized for high-throughput processing on NVIDIA hardware using vLLM.
| Parameter | Before | Now |
|---|---|---|
| Languages (Transcription) | Not specified | 40 |
| Translation support | Not specified | 31 languages to/from English |
| Speaker attribution | Not specified | 13 languages (up to 8 speakers) |
Why it matters
This model consolidates transcription, diarization, and translation into a single pipeline, significantly reducing complexity for contact-center applications. The promptable nature allows developers to enforce domain-specific formatting (e.g., medical or legal) without complex post-processing.
What it means for AI agents and contact centers
This model is a strong candidate for replacing multi-step voice processing pipelines. Benchmark performance on diarization and long-form transcription suggests potential improvements for your analytics and QA tools. When testing, prioritize comparing diarization accuracy and latency against current solutions, specifically focusing on its ability to handle domain-specific terminology via custom prompts.
🧪 Worth Testing
The model's efficiency and ability to handle complex speaker attribution make it a high-priority candidate for improving voice-agent accuracy and diarization performance.
Sources
- Hugging FaceOfficialPrimary sourceOriginal article →„NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt“28 Oct 2026, 16:58
- Published by source
- 28 Oct 2026, 16:58
- Found by our system
- 28 Oct 2026, 17:22
- Summary generated
- 28 Oct 2026, 17:22
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy