I agree that making web pages takes a lot of testing on different browsers and OSes. And that, as you said, there is no write-once-run-anywhere in software. But

I agree that making web pages takes a lot of testing on different browsers and OSes. And that, as you said, there is no write-once-run-anywhere in software. But I think we should minimize the problem, not make it worse. Having the same JS TTS gets you much closer to the ideal, using system TTS gets you farther. I do agree that a native library will have better performance, I am estimating something like 3-5X faster in the near future. So that is a benefit to the system TTS approach. But even the fairly unoptimized version in this online demo isn't too slow to be useful, I don't think, and it can be made much faster if necessary. I haven't focused on speed yet. My concern with each OS having its own voices is that the names of the voices aren't enough to know what your users will hear. Unless we have a standard for TTS that includes the actual voice data, otherwise say "male UK English" may sound very different on different platforms. It's clear there are tradeoffs here, both ways, and you make sound points. But I prefer the JS TTS approach, unless you are writing something like an iPad-specific app. If you don't care about other platforms, then I agree, system TTS is better.