2 ms·

Very cool project and good job on releasing a new version! Libraries like these are huge work. One question: compared to the native text-to-speech on macOS, th

Very cool project and good job on releasing a new version! Libraries like these are huge work. One question: compared to the native text-to-speech on macOS, the synthesized speech sounds, for a lack of a better word, robotic. Is this an inherent property of the approach you used or a result of trying to squeeze something as complex as this into a Javascript library?

masswerk 7y ago
This is based on eSpeak [1] for Un*x environments, which is based on an application for Acorn/Risc OS. So, yes, it's quite dated. On the other hand, it's lean enough to be run in realtime in Emscripten... However, all the configuration data, including phoneme tables, may be overwritten (but you would have to install eSpeak on your machine first, in order to compile these.) Another approach would be actually porting this to JS (instead of cross-compiling), by this having full access. But I simply do not have the resources for this. (Meanwhile, there's the Web Speech Synthesis API. With this being available on most modern clients, it's probably not worth the effort.) [1] http://espeak.sourceforge.net/ http://espeak.sourceforge.net/