aws-polly

very low latency amazon Text-To-Speech aws polly script.
a simple to use alternative, but with a ~3s execution time, is listed further down.
this sctipt runs as daemon and waits for the bash SIGUSR1 signal, to achieve a very low playback latency time (under 0.5s).

Installation

1. download project to /usr/local/aws-polly/

2. install dependency:

sudo apt install mpg123 

3. modules install

npm i aws-sdk dotenv fs child_process minimist

4. rename file .env-blank to .env

5. add your aws credentials to .env

Usage

call the following from your app / script as root:

pkill -x -o -SIGUSR1 -f 'node /usr/local/aws-polly/signal.js' || node /usr/local/aws-polly/signal.js &

this will:

• send a SIGUSR1 signal to the script,
• or start it in background if it doesn't run.

run node signal.js --help for information about the optional cli arguments, and add them (twice) to the above line.
by default, signal.js will load /tmp/tts.txt, read it using voice-id 'Matthew', and play it on the system default speaker.

list pid of running instance:

pgrep -f 'node /usr/local/aws-polly/signal.js'

kill running instance:

pkill -SIGINT -f "node /usr/local/aws-polly/signal.js"

PSA:

• all arguments are necessary twice, because otherwise pkill will match its own child_process.exec() process when executed from node.js

Usage for export module function.js (outdated and missing features!!)

const POLLY = require("/usr/local/aws-polly/function.js");

call from your node code:

POLLY.ttsBurrito();

References

script is based on instructions and examples from:
https://medium.com/@anaptfox/getting-started-with-amazon-polly-using-node-js-345e84dbd23d
https://medium.com/better-programming/text-to-speech-build-apps-that-talk-with-aws-polly-and-node-js-a9cdab99af04
https://trevorsullivan.net/2016/12/01/amazon-aws-cloud-polly-nodejs/

Listen to some voice examples:
https://www.amazonaws.cn/en/polly/
Listen to all voices (requires aws console login) :
https://us-east-2.console.aws.amazon.com/polly/home/SynthesizeSpeech
Full list of available voices: (write voice names without phonetics. Léa -> Lea)
https://docs.aws.amazon.com/polly/latest/dg/voicelist.html

check how many characters you've used this month here: https://console.aws.amazon.com/billing/home#/bills

SSML (markup language) reference doc:
https://developer.amazon.com/it-IT/docs/alexa/custom-skills/speech-synthesis-markup-language-ssml-reference.html#prosody
AWS polly class properties doc:
https://docs.aws.amazon.com/AWSJavaScriptSDK/latest/AWS/Polly.html

CLI alternative

an easy to use alternative are the CLI packages from aws themselves. execution time until playback is about 2.7s.
a complete command could then look like this:

aws polly synthesize-speech --voice-id Matthew --output-format mp3 --text "<speak>$(cat /tmp/tts.txt)</speak>" --text-type ssml /tmp/tts.mp3 && mpg123 -q --no-control -a plughw:1 /tmp/tts.mp3 &

(~5 simultaneous instances will result in failure)
command reference:
https://awscli.amazonaws.com/v2/documentation/api/latest/reference/polly/synthesize-speech.html
CLI v2 install:
https://docs.aws.amazon.com/cli/latest/userguide/install-cliv2-linux.html
CLI v2 is not supported on raspberry pi, but CLI v1 is still available for arm processors.
CLI v1 install:
https://docs.aws.amazon.com/cli/latest/userguide/install-linux.html#install-linux-bundled-sudo
either create the aws credentials files or run aws config
https://blog.gruntwork.io/authenticating-to-aws-with-the-credentials-file-d16c0fbcbf9e?gi=f46ca71979f8

ignore this part. pkg workflow notes for future self

0. copy creds from old .js
1. cp signal.js aws-polly-signed-signal.js
2. add creds into param
3. npm i pkg
4. get https://github.com/vercel/pkg-fetch , or for armv7 https://github.com/robertsLando/pkg-binaries
5. pkg --targets node10.15.3-linux-armv7 package.json
6. mv aws-polly-signed-signal ../