leepro / alexa-polly

Webservice to transform Polly's output in a format accepted by Alexa SSML

AWS Polly to Alexa MP3 converter

This is a web service that generates TTS MP3 using AWS Polly and prepare them to use with Alexa. This web service is implemented as an AWS Lambda function, bundled with ffmpeg's encoding tool. The code is deployed using AWS SAM serverless framework.

Usage

$ curl https://o4gxgxfy8k.execute-api.us-east-1.amazonaws.com/Prod/polly/Celine -d "Bonjour"
<audio src="https://alexademo.ninja/polly/c657a7e1-1fc2-4f79-93b8-635abfaeba16"/>

You can test the generated MP3 by copy / pasting the code in the Alexa developer console.

Aletrnatively, on MAc, you can type the following commands :

$ wget -q https://alexademo.ninja/polly/c657a7e1-1fc2-4f79-93b8-635abfaeba16
$ afplay c657a7e1-1fc2-4f79-93b8-635abfaeba16
$ rm c657a7e1-1fc2-4f79-93b8-635abfaeba16

Pre-requesites

AWS CLI

Obviosuly, you need an AWS Account :-) and AWS CLI properly configured

S3 bucket

You need to have an S3 bucket in the same region as the one where you will deploy the code. That bucket must be configured to serve HTTP requests.

To make the bucket available as https://<bucket_name> with a TLS certificate accepted by Alexa,
I exposed the bucket through a CloudFront distribution, my own domain name and a certificate
provided by AWS Certificate Manager.

Deploy

1. Modify the name of the bucket in package.sh

2. In a terminal, type

./package.sh && ./deploy.sh

TODO

Expose the API endpoint with a custom domain name and TLS certificate