Audio2Face-3D

Getting Started

• Overview
• Getting Started
  • Prerequisites
  • Launching the Audio2Face-3D NIM

Architecture

• Architecture overview
  • Audio2Face-3D NIM Data Flow
• Audio2Face-3D Microservice
  • Bidirectional gRPC Communication
  • Configuration
• Observability
  • Overview
  • Traces
  • Metrics
  • Quick start the NIM container and visualize metrics and traces

Interacting with Audio2Face-3D

• Sample application connecting to Audio2Face-3D
  • Prerequisites
  • Setting up the sample application
  • Health checking
  • Interacting with Audio2Face-3D
  • Results
  • Resampling Audio Data
• Performance
  • Performance evaluation
  • Modifying deployment configurations for benchmarking
  • Results
  • Performance benchmark
  • Performance troubleshooting
• Fetch current configuration from Audio2Face-3D
  • Assumptions
  • Setting up the config fetcher application
  • Fetching configuration
  • Results
• gRPC with Audio2Face-3D
  • Bidirectional Streaming gRPC
  • Health Check gRPC
  • Config Fetch gRPC
• Unreal Engine Interaction
  • Using Unreal Engine with Audio2Face-3D

Deployment

• Audio2Face-3D NIM Container Deployment and Configuration Guide
  • Prerequisites
  • Configuration files
  • Configuration Usage
  • Model Cache Management
  • Launching A2F-3D NIM with Custom Entrypoint
  • Streamlined Configuration Updates
  • Flexible Configuration Management
  • Advanced Stylization
  • Configuration files for Unreal Engine Metahuman
  • Parameter Tuning Guide
  • Environment variables
  • Volumes
  • Quick Deployment of Audio2Face-3D Microservices
• TLS Support
  • Certificate Generation Process
  • Deploying Audio2Face-3D NIM with TLS
  • Sample Applications
• One Click Deployment
  • Overview
  • Supported Platforms
  • System Requirements
  • Deployment Package Structure
  • Deployment Process
• Kubernetes Deployment
  • 1. Install microk8s, GPU addon, set local path provisioner and set secrets
  • 2. Download and install the helm chart
  • 3. Optional - Expose the pod IP publicly
  • 4. Check it is up and running
  • Uninstall the app

Notes and changelog

• Support Matrix
  • Models
  • Optimized configurations
  • Non-optimized configurations
  • Software
• Migration Guide
  • Migrating from 1.3 to 2.0
  • Migrating from 1.2 to 1.3
  • Migrating from 1.0 to 1.2
  • Audio2Face Controller
  • Audio2Face-3D Microservice
• Sharing Audio2Face-3D Compute Resources
  • Sharing Audio2Face-3D compute with a Renderer
• Network Customization and Accessibility
  • Audio2Face-3D Docker-Compose Customization
  • Exposing Audio2Face-3D Deployment from WSL
• Glossary of Terms
  • Common terminology
  • NVIDIA technology
  • 3rd Party Technology
• Security
  • Notes
• Release Notes
  • v2.0
  • v1.3.16
  • v1.3.15
  • v1.2.0
  • v1.0.0
• Troubleshooting
  • Known issues
  • Lip Sync Issues with Burst Mode (Tokkio)
  • Q&A
  • Reducing VRAM Usage for Edge Deployments
  • gRPC Response Status Code
  • Audio2Face-3D NIM Public API
  • GPU Device ID Mismatch Warning (nim_list_model_profiles)
  • TensorRT Engine Build Fails for Diffusion Models (Insufficient Memory)
  • CUDA Error 802 “system not yet initialized” on B200 Multi-GPU Systems
  • Sample Scripts Fail with “Connection refused” Traceback
  • Invalid Model Name Silently Falls Back to Default (james_v2.3.1)
  • Cached TRT Engines Don’t Match Requested Model
  • NIM Fails When Using a Profile for a Different GPU
  • Permission Denied on Cache Directory (/tmp/a2x)
  • CUDA Error: No CUDA-Capable Device Detected
  • Stopping and Restarting the A2F Server
  • Kubernetes Pod Stuck in Pending Status
  • kubectl Connection Refused on localhost:8080
  • MicroK8s Fails to Start (Containerd Config Version Mismatch)
  • Snap Tools Fail with “Home Directories Outside of /home”
  • Kubernetes Pod Fails with “CUDA-capable device(s) is/are busy or unavailable”