# NVIDIA-NeMo/Speech

A scalable generative AI framework built for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech)

Repository: https://github.com/NVIDIA-NeMo/Speech
Canonical: https://ross.abutalabs.com/products/speech
Homepage: https://docs.nvidia.com/nemo/speech/nightly/index.html
Language: Python
License: Apache-2.0
License Family: permissive
Topics: machine-translation, speaker-recognition, asr, tts, generative-ai, deeplearning, neural-networks, speaker-diariazation, speech-translation, speech-synthesis
Last push: 2026-08-27T00:01:30+00:00

## Health v2 (maintenance only)
Score: 98/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 96, longevity 100
- inputs: {"age_days": 2585, "days_push": 7, "days_rel": 27, "gap_med": 19, "n_releases_24m": 24}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 18337, forks 3595 (observed 2026-08-28T04:11:26.586640+00:00)

## What it is
NVIDIA NeMo Speech is an open-source Python framework for building, training, and deploying speech, audio, and multimodal language models, covering ASR, TTS, speaker tasks, and speech-aware LLMs. It ships production-ready pretrained checkpoints, modular neural components, and scalable multi-GPU/multi-node training via PyTorch Lightning with Hydra-based YAML configuration.

## Use cases
- transcribe audio to text with streaming asr
- convert text to natural speech with tts
- diarize who spoke when in multi-speaker audio
- fine-tune speech recognition models on custom data
- build speech-aware large language models
- enhance and separate audio signals
- train tts models on multiple languages
- run speaker recognition and verification

## When to choose
- you need state-of-the-art pretrained ASR or TTS models with a path to production
- you want scalable multi-GPU/multi-node training for speech models
- you need streaming ASR with controllable latency
- you are researching speech LLMs or multimodal audio models
- you want speaker diarization, recognition, and verification in one toolkit

## When to avoid
- you only need simple audio playback or editing rather than AI models
- you have no GPU and need fast inference (CPU-only is slow)
- you want a lightweight single-purpose ASR library without a large dependency stack
- you need non-PyTorch frameworks like TensorFlow or JAX

## Facets
- artifact type: framework
- maturity: active
- function: speech-recognition, tts, machine-learning, deep-learning, audio-processing, llm-training, sdk
- domain: speech-processing, machine-learning, deep-learning, artificial-intelligence
- platform: python, cloud
- tags: asr, tts, speaker-diarization, speech-translation, pytorch-lightning, nvidia, pretrained-models, speech-llm, voice-activity-detection, forced-alignment, audio, natural-language-processing, gpu, linux, docker

## Member repositories
- NVIDIA-NeMo/Speech (main) score 98

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:26.586640+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:02:04.387459+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA-NeMo/Speech (fetched 2026-08-28T04:11:26.586640+00:00, sha 7542fca65f19)
  - homepage: https://docs.nvidia.com/nemo/speech/nightly/index.html (fetched 2026-08-29T07:59:55.409534+00:00, sha 88940b7bbfb3)
  - site_page: https://docs.nvidia.com/nemo/speech/nightly/starthere/install.html (fetched 2026-08-29T07:59:55.454701+00:00, sha 01e754bc554d)
  - site_page: https://docs.nvidia.com/nemo/speech/nightly/features/parallelisms.html (fetched 2026-08-29T07:59:55.486782+00:00, sha ffe16f442542)
  - site_page: https://docs.nvidia.com/nemo/speech/nightly/features/mixed_precision.html (fetched 2026-08-29T07:59:55.524770+00:00, sha a27ec8c772f6)
  - site_page: https://docs.nvidia.com/nemo/speech/nightly/asr/speaker_diarization/resources.html (fetched 2026-08-29T07:59:55.561362+00:00, sha b5e462cc79f8)
  - site_page: https://docs.nvidia.com/nemo/speech/nightly/asr/speaker_recognition/resources.html (fetched 2026-08-29T07:59:55.597965+00:00, sha fe734b54483c)
  - site_page: https://docs.nvidia.com/nemo/speech/nightly/asr/ssl/resources.html (fetched 2026-08-29T07:59:55.642323+00:00, sha b8ad5cdef481)
  - site_page: https://docs.nvidia.com/nemo/speech/nightly/asr/speech_classification/resources.html (fetched 2026-08-29T07:59:55.669136+00:00, sha 42abd7fde7af)
  - site_page: https://docs.nvidia.com/nemo/speech/nightly/tts/intro.html (fetched 2026-08-29T07:59:55.685245+00:00, sha 3338cc790bf9)
- Data as of 2026-08-30T08:39:29.467469+00:00.