NVIDIA NeMo Speech Developer Docs#

NVIDIA NeMo Speech is an open-source toolkit for speech, audio, and multimodal language model research, with a clear path from experimentation to production deployment.

Important

New: Nemotron 3.5 Transcribe — offline transcription in 40 languages, with instruction control, vocabulary context, speaker attribution, word timestamps, speech translation, and audio question answering. See the quickstart for installation and your first transcription.

What is NeMo Speech?#

NVIDIA NeMo Speech is an open-source toolkit for building, customizing, and deploying speech, audio, and multimodal language models. It provides:

  • Pretrained models — production-ready checkpoints on NGC and HuggingFace Hub

  • Modular architecture — neural modules you can mix, match, and extend

  • Scalable training — multi-GPU/multi-node via PyTorch Lightning with mixed-precision support

  • Simple configuration — YAML-based experiment configs with Hydra

Get started (install the PyTorch build for your platform first):

uv pip install 'nemo-toolkit[asr,tts]'
import nemo.collections.asr as nemo_asr
model = nemo_asr.models.ASRModel.from_pretrained("nvidia/parakeet-tdt-0.6b-v2")
print(model.transcribe(["audio.wav"])[0].text)

Trying to finetune a model?#

Check out our latest /nemo-speech-finetune-asr agent skill.