Without modifying Chromium, I do not think you can add or customize voices.

Did you acquire and set API keys (see https://www.chromium.org/developers/how-tos/api-keys for instructions)? Without them, speech synthesis does not work.
The speech synthesis feature of Chrome/Chromium uses a remote Google API, as far as I know, so unless you are allowed to use that API (you have an API key and you have not exceeded the usage quota), this will not work.

Able to reproduce this issue on reported version 60.0.3112.78, on stable version 61.0.3163.100 and on latest dev 63.0.3223.8 using Ubuntu 14.04,Windows 10 and Mac 10.12.6 using attached .html file. i.e; First list shows 0 and later shows list count. Attaching screenshot of same.

This issue is seen from M50 checked on 50.0.2652.0, so considering it as non-regression and marking it as Untriaged.

Deleted: Issue 768186.png 176 KB

	Issue 768186.png 176 KB View Download

Thanks by the information phistuck!

You mean a Chromium user must follow all this steps to use a TTS web app? :-(

Is this a temporary requirement? Could Chromium project use the same Firefox voices to not conflict with Google's service rules?

Yes, a Chromium user must follow those steps (unless someone had already followed them for them and embedded the API keys in their customized binary, like Chrome does). Remember that Chromium is not meant to be used for general browsing and that the Chromium project does not release official Chromium binaries.
Whoever releases those binaries can choose to embed API keys (though it is somewhat illegal, as most of them are not even available for licensing and are only meant for Chrome/Google products. I guess speech synthesis is one of them).

I guess that if someone contributed a lightweight speech synthesis adapter implementation that makes Chromium use other voices, the project might accept it, but I am not a Googler, so I am just speculating here.

And this is not a temporary requirement, as far as I understand.

Have you tried attaching onvoiceschanged event handler to window.speechSynthesis, then checking if .getVoices() returns an array of voices?

.getVoices() returns results asynchronously, use onvoiceschanged event

window.speechSynthesis.onvoiceschanged = e => {
  const voices = window.speechSynthesis.getVoices();
  console.log(voices);
}
window.speechSynthesis.getVoices();