skills.rest
NVIDIA
NVIDIAOFFICIAL
Explore all skills by this author →
WebsiteGitHub

nemotron-speech

Deploys, runs, and tests NVIDIA Riva ASR, TTS, and NMT NIMs on cloud or self-hosted infrastructure.

3.2k|370|Updated Feb 25, 2026
Search Tags:#text-to-speech#Deployment#speech-to-text#asr#NVIDIA#NIM#riva
One-click install
npx skills add https://github.com/NVIDIA/skills --skill nemotron-speech
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: nemotron-speech
Source: https://github.com/NVIDIA/skills/tree/main/skills/nemotron-speech
Command: npx skills add https://github.com/NVIDIA/skills --skill nemotron-speech

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires nvidia-riva-client, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Deploying and operating NVIDIA Nemotron Speech (Riva) NIMs involves scattered documentation, per-release model catalogs, container configuration, and protocol choices that are easy to get wrong. This Skill routes any speech AI task — speech-to-text, text-to-speech, or translation — to the correct deployment and inference workflow with current, verified commands.

Core Features & Use Cases

  • ASR, TTS, and NMT deployment: Deploy Parakeet, Canary, Whisper, and Magpie models via build.nvidia.com cloud inference or self-hosted Docker NIMs with gRPC, HTTP, and WebSocket clients.
  • Custom model pipelines: Convert fine-tuned NeMo checkpoints (.nemo) to deployable NIMs through the riva-build and riva-deploy RMIR workflow.
  • Pipeline tuning and readiness: Configure VAD, diarization, language models, SSML, zero-shot voice cloning, and IPA pronunciation dictionaries, plus run 6-step GPU/driver/NGC readiness checks.
  • Use Case: Ask your agent to "deploy a self-hosted Parakeet ASR NIM and transcribe a WAV file over gRPC" and it walks through container launch, health verification, and inference step by step.

Quick Start

Ask your agent to deploy a Riva ASR NIM for Parakeet and transcribe an audio file, and it will route to the right workflow automatically.

Frequently Asked Questions about nemotron-speech

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I deploy a Riva ASR NIM for speech-to-text?▼

Set CONTAINER_ID and NIM_TAGS_SELECTOR from the ASR support matrix, run the Docker container with GPU access and an NGC API key, verify readiness via the /v1/health/ready endpoint, then transcribe with the nvidia-riva-client over gRPC on port 50051.

How do I convert a fine-tuned NeMo checkpoint to a Riva NIM?▼

Run riva-build inside the NIM container with an inline nemo2riva config block matching your model family to produce an RMIR, then run riva-deploy to build the Triton model repository, and launch the NIM with NIM_DISABLE_MODEL_DOWNLOAD=true.

Can I use Riva speech models without a GPU?▼

Yes, cloud-hosted inference via build.nvidia.com requires only the nvidia-riva-client pip package and an NVIDIA_API_KEY. You resolve the model's function ID from the NVCF Functions API and call grpc.nvcf.nvidia.com:443 with SSL.

How do I fix mispronounced words in Riva TTS output?▼

Generate IPA pronunciation candidates for the word, synthesize test variants using the custom_dictionary field with a double-space delimiter, pick the best one, then apply it via custom_dictionary wire format, SSML phoneme tags, or a session dictionary file.

Why does my Riva NIM container fail to become ready?▼

Common causes include a wrong NIM_TAGS_SELECTOR, insufficient VRAM for the model, missing NGC authentication, or cache directory ownership issues. Run the 6-step readiness check covering architecture, driver, compute capability, VRAM, Container Toolkit, and NGC credentials.

What are the limitations of self-hosted Riva NIMs?▼

Self-hosted deployment requires x86_64 Linux, an NVIDIA AI Enterprise license, and a supported GPU with sufficient VRAM. WSL2 supports only a subset of NIMs via Podman, and audio input must be mono WAV (16-bit PCM) or Opus.

Related Skills

Explore More Software Engineering Skills→
jayll1303jayll1303
17

docker-gpu-setup

Build GPU-enabled Docker containers with NGC base images and uv.

Community
Advanced
J4JoshuaJ4Joshua

livekit-agents

Orchestrate offline realtime voice pipelines with local STT, LLM, and TTS.

Community
Advanced
castrojocastrojo

transcript-deep-analysis

Extract architectural patterns and CNCF project integrations from video transcripts.

Community
Advanced
marvelousempiremarvelousempire

kokoro-fastapi

Integrate and manage Kokoro FastAPI for text-to-speech with ONNX and GPU PyTorch.

Community
Intermediate