Support Matrix

The following tables list the model, languages, supported GPUs, software prerequisites, and host resource requirements for the LipSync NIM.

Models

The following table lists the model packaged in this NIM.

Model Name	Model ID	Publisher
LipSync	lipsync	NVIDIA

Supported Languages

The LipSync NIM provides a generic, language-agnostic model (default) and language-specific fine-tuned models. Select a model at deployment time with the NIM_TAGS_SELECTOR environment variable; refer to Selecting a Language-Specific Model.

Language	Code	Model
Language-agnostic (default)	(none)	Generic
German	de	Fine-tuned
Spanish	es	Fine-tuned
French	fr	Fine-tuned

Optimized Configurations

The following server GPUs are optimized for the LipSync NIM:

Server GPU	Precision
T4	FP16
A2, A10, A16, A40	FP16
L4, L40, L40s	FP16
RTX PRO 6000 Blackwell Server Edition	FP16
RTX PRO 4500 Blackwell Server Edition	FP16

Other architectures (Consumer RTX GPUs)

The following consumer RTX GPUs are also supported at FP16 precision.

Consumer GPU	Precision
RTX 4090	FP16
RTX 5090, 5080	FP16

The LipSync NIM is compatible with professional and consumer GPUs that have Tensor cores and are based on the following NVIDIA architectures: Blackwell, Ada, Ampere, and Turing. The RTX-based GPUs are also supported.

The NIM requires NVENC/NVDEC hardware. GPUs without NVENC/NVDEC hardware support are not supported, including A100, H100, and B100 products. For details about supported GPUs and H264 YUV formats, refer to the Video Encode and Decode GPU Support Matrix.

Software

The LipSync NIM requires the following host software, NVIDIA platform components, and disk and memory resources.

NVIDIA Driver and Prerequisites

NVIDIA driver requirements and other prerequisites for LipSync NIM:

Prerequisite	Version	Download and install steps
NVIDIA Graphic Drivers for Linux	571.21+	https://www.nvidia.com/en-us/drivers/unix/
Docker	latest	Ubuntu, CentOS, and Debian: https://docs.docker.com/engine/install/; Rocky Linux: https://docs.rockylinux.org/gemstones/containers/docker/
NVIDIA Container Toolkit	latest	Installation and configuration instructions: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html
Python (sample client only)	3.12+	https://www.python.org/downloads/. A virtual environment is required.

On bare-metal Linux systems with PCIe topology and multiple GPUs, we recommend setting IOMMU to passthrough. Refer to PCIe Multi-GPU Systems.

LipSync NIM uses the following NVIDIA software platforms:

Components	Version
CUDA	12.8.1
cuDNN	9.7.1.26
TRT	10.9.0.34
Triton Inference Server	v2.56.0
DeepStream	8.0

Host Resource Requirements

Approximate figures for planning. Actual usage varies with input resolution, concurrency, and the selected model.

Resource	Requirement
Container image	Approximately 9.5 GB download
Model artifacts	Downloaded on first start into NIM_CACHE_PATH. Mount a persistent volume to avoid re-downloading on every launch.
Disk	Set aside disk space for the image, model artifacts, and your input and output media.
Shared memory	--shm-size=8GB, as shown in every launch command.
GPU	One GPU with NVENC/NVDEC hardware. Concurrency above 1 increases GPU memory use; see NV_AI4M_MAX_CONCURRENCY_PER_GPU.