How to deploy TTS containers

Azure speech containers offers TTS and SR containers to run in on-premise machines.

The document above describes the download, run the containers using docker command.

In a real production environment, usually you need to have a container orchestration like Kubernetes (k8s). So it can automate the deployment, scaling and management of the containers.

If your company already has infrastructure to run k8s and deploy containers, skip to step 3 and 4.

Step 1: install k8s

There are many ways to do it. Here are a few links:

• Ubuntu K8s: https://ubuntu.com/kubernetes/install
• Azure stack: it is possibly to run k8s on Azure stack using aksengine

Step 2: setup helm and make sure you can use it to install container

Here are some useful docs: https://www.digitalocean.com/community/tutorials/how-to-install-software-on-kubernetes-clusters-with-the-helm-package-manager

Step 3: install TTS containers

To run speech containers in best performance, it is recommended to allocate K8S node with equal or higher configuration than what we uses in Azure cloud.

After node is setup , it should be staightforward to deploy Pods if step 2 is ok. You can set the # of PODs for the containers to run and scale.

Step 4: test it with speech SDK

Connect to the deployed container service using the IP address or local service url.