MLX-compatible version of nvidia/nemotron-3.5-asr-streaming-0.6b, prepared for use with OpenVox and other Apple Silicon workflows using mlx-audio.

Nemotron 3.5 ASR Streaming is NVIDIA's 600M-parameter cache-aware streaming FastConformer-RNNT speech recognition model with language-ID prompt conditioning.

It supports 40 language-locales along with punctuation and capitalization.

This repository contains bfloat16 weights for full-quality inference.

For architecture details, benchmarks, training information, limitations, and intended use, see the original NVIDIA model card.

OpenVox

This model is distributed through the theoracleguy Hugging Face repository for use with OpenVox, a local AI voice platform for Mac, iPad, and Windows.

OpenVox is designed around local AI processing, allowing supported speech and voice workflows to run directly on your device.

This repository can also be used independently with mlx-audio.

Install

This model requires mlx-audio with Nemotron ASR support.

If your installed PyPI version does not yet include Nemotron support, install the latest version directly from GitHub:

pip install -U "git+https://github.com/Blaizzy/mlx-audio.git"