Intelligent Voice is pleased to announce the 6.1.1 release. Full details can be found in the Intelligent Voice 6.1.1 release note.
All IV 6.1.1 components can be installed using the latest version 4.0 of the IV universal installer:
The installer is an Ansible playbook which installs IV components onto one or more Red Hat, Oracle Linux or Ubuntu servers. The installer itself can run on any Linux platform with Ansible 2.14+ and OpenSSL 3.0.2+, and can install the IV software onto the same server or remotely using SSH.
Internet access requirements for installation
By default the installer will connect to the internet to download the required components. All sources are configurable, so the installer can run on a server without any internet access if the requirements below are mirrored locally.
Once installation is complete, no internet access is required.
1. Pre-requisite packages
The installer will attempt to download and install all pre-requisite packages if not already present. If all the pre-requisite packages are already present, no action is taken.
These will be installed using yum, dnf, apt or pip as appropriate, using existing configured package repositories and adding two new repositories: Docker (from
download.docker.com) and NVIDIA (from
The pre-requisite packages for all supported operating systems are:
requestspackages (installed with pip)
nvidia-container-runtimefor GPU support (if you set
Additional pre-requisite packages for Red Hat and Oracle Linux are:
2. Container images
Container images will be downloaded from:
IV’s container registry at http://intelligentvoice.azurecr.io
Docker Hub (http://registry.hub.docker.com )
NVIDIA NGC (nvcr.io)
See the README.md file included in the installer for details on configuring a local private registry, which will remove the need for internet access during installation.
3. IV models
IV models will be downloaded from models.intelligentvoice.com. To remove the requirement for internet access, set the
download_models option to false or change the model URLs to point to a local mirror of the models.