Is this page helpful?

Introduction#

NVIDIA NeMo Toolkit is an open-source toolkit for building, customizing, and deploying speech, audio, and multimodal language models. It allows for the creation of state-of-the-art models for speech and audio processing. For detailed information on utilizing NeMo in your generative AI workflows, refer to the NeMo Framework User Guide.

Training generative AI architectures typically requires significant data and computing resources. NeMo utilizes PyTorch Lightning for efficient and performant multi-GPU/multi-node mixed-precision training. For Speech AI applications, Automatic Speech Recognition (ASR) and Text-to-Speech (TTS), NeMo is developed with native PyTorch and PyTorch Lightning, ensuring seamless integration and ease of use.

NVIDIA NeMo Toolkit features separate collections for Automatic Speech Recognition (ASR), Text-to-Speech (TTS), Audio Processing, and SpeechLM2 models. Each collection comprises prebuilt modules that include everything needed to train on your data. These modules can be easily customized, extended, and composed to create new generative AI model architectures.

Pre-trained NeMo models are available to download on NGC and HuggingFace Hub.

Prerequisites#

Before using NeMo, make sure you meet the following prerequisites:

  1. Python version 3.12 or above.

  2. PyTorch version 2.7+.

  3. Access to an NVIDIA GPU for model training.

Installation#

From PyPI:

pip install nemo_toolkit[asr,tts]

Available extras: asr, tts, audio, common.

From source:

git clone https://github.com/NVIDIA/NeMo.git
cd NeMo
pip install -e '.[asr,tts]'

Quick Start Guide#

To explore NeMo’s capabilities, here are examples for ASR, TTS, speaker diarization, and speech language models.

ASR with Parakeet#

import nemo.collections.asr as nemo_asr

asr_model = nemo_asr.models.ASRModel.from_pretrained(model_name="nvidia/parakeet-tdt-0.6b-v2")
transcript = asr_model.transcribe(["audio.wav"])
print(transcript)

Text-to-Speech#

import nemo.collections.tts as nemo_tts

spectrogram_generator = nemo_tts.models.FastPitchModel.from_pretrained(model_name="tts_en_fastpitch")
vocoder = nemo_tts.models.HifiGanModel.from_pretrained(model_name="tts_en_hifigan")

parsed_text = spectrogram_generator.parse("Hello, welcome to NeMo!")
spectrogram = spectrogram_generator.generate_spectrogram(tokens=parsed_text)
audio = vocoder.convert_spectrogram_to_audio(spec=spectrogram)

import soundfile as sf
sf.write("output.wav", audio.to('cpu').detach().numpy()[0], 22050)

Speaker Diarization with Sortformer#

from nemo.collections.asr.models import SortformerEncLabelModel

diar_model = SortformerEncLabelModel.from_pretrained("nvidia/diar_streaming_sortformer_4spk-v2")
diar_model.eval()
predicted_segments = diar_model.diarize(audio=["meeting.wav"], batch_size=1)
for segment in predicted_segments[0]:
    print(segment)  # begin_seconds, end_seconds, speaker_index

Speech-to-Text with Canary-Qwen (SpeechLM2/SALM)#

from nemo.collections.speechlm2.models import SALM

model = SALM.from_pretrained('nvidia/canary-qwen-2.5b')
answer_ids = model.generate(
    prompts=[[{
        "role": "user",
        "content": f"Transcribe the following: {model.audio_locator_tag}",
        "audio": ["speech.wav"],
    }]],
    max_new_tokens=128,
)
print(model.tokenizer.ids_to_text(answer_ids[0].cpu()))

For detailed tutorials and documentation on specific tasks or to learn more about NeMo, check out the NeMo tutorials or dive deeper into the documentation, such as learning about ASR in here.

Discussion Board#

For additional information and questions, visit the NVIDIA NeMo Discussion Board.

Contribute to NeMo#

Community contributions are welcome! See the CONTRIBUTING.md file for how to contribute.

License#

NeMo is released under the Apache 2.0 license.