Hosting and Implementing Audio2Face REST API on a Remote Ubuntu Server

[jimi_12]

Hello NVIDIA Community,

I am currently developing an API that processes audio input using the Audio2Face REST API. I am working on a remote Ubuntu 22.04 server equipped with an NVIDIA RTX GPU. Here are the GPU details:

I’m currently accessing the server via SSH, and Docker is installed on the server. I’m using a devcontainer and building the environment through a Dockerfile.

My goal is to separate vocals from the input audio, pass the .wav file to Audio2Face, and receive blendshape outputs via the REST API. However, I am encountering challenges in determining where and how to host the A2F server. Initially, I planned to use the Omniverse Launcher AppImage for Linux, but found out it does not support GUI based on forum results. Now, I am unsure whether to use the Omniverse SDK, install the A2F app within it, or use ACE as some forums have suggested.

I am slightly confused and seeking precise information on how to proceed. Any detailed guidance would be greatly appreciated.

Thank you!

[Richard3D]

Non of the Omniverse or the A2F (which is not part of OV) supports SSH. It REQUIRES a GUI. These are powerful graphical applications. Yes we support Dockers, containers, Windows, Linux and Python coding, but not through SSH. You need to use a full Remote Desktop with GPU passthrough support and the right grid drivers.

[jimi_12]

Then ACE is the sole solution for this?

[Richard3D]

ACE is also very intensive and relies on a powerful GUI. The solution is as above. Keep your workflow, but just don’t use SSH. Use a GUI system. Just use VNC or such. Why do you have to use a text based command system for visual software.