[Example Request] Is it possible to use SageMaker for streaming speech recognition?

Hi @javier @tomfaulhaber @channy @danny @ishaaq ,
We are working on a custom speech-to-text project, and we need to deploy the real-time speech-to-text server in the cloud. While checking the SageMaker, we found few limitations like 60 seconds timeout. So we have some doubts regarding this.
For real-time speech to text (audio-stream to text), can we use SageMaker?
Using Lambda + ApiGateway (WebSocket) for audio-stream and pass the audio to SageMaker. Is it a good solution?

Other than SageMaker, we are considering using the ECS cluster with auto-scaling, Will it take time to initialize the new cluster while auto-scaling?.

If anybody hosted a real-time speech recognition server on AWS, please help us decide on a solution.

Let us know if there is any better stack for real-time speech recognition stack we can use in AWS.

The issue discusses SageMaker, Lambda, API Gateway WebSocket, and ECS for real-time speech recognition, but names no repository files, tests, or entry points. First clarify whether an AWS example is requested and which architecture should be demonstrated; the issue provides no concrete completion criterion.

Written by the indexing model from the issue text.