Note by @fastfinge

There will be people throwing things at their screen or braille display when I say this, but I have no idea what the difference is between IBMTTS and Eloquence. If there's a sonic difference, can't hear it. Never grew up with it, so I just don't get it.
School me, go on.

@FreakyFwoof If you’re talking about the NVDA addons, the IBMTTS addon usually uses the eloquence DLL files, so the sound is identical. The differences between the two addons are features and what comes included. If you mean the difference between Eloquence and IBMTTS/ViaVoice, Eloquence licensed there code to IBM. IBM went on to add some new features, and many new and exciting bugs. And also change the API for no particular reason. And created a new dictionary format that’s neither better or worse, just different for no explicable reason. Also, the IBMTTS dll files can be had from IBM’s FTP site. The Eloquence dlls have to be ripped out of commercial software. Also, Apple and Nokia had completely different eloquence versions that are 99 percent the same, but each have subtle differences. And as far as I can tell, all of the voice and dictionary rules were created in a custom programming language, and the source code for the rules, and the compiler that compiles them, is almost certainly lost. That’s why Apple, for example, includes the community dictionary rather than just updating eloquence to fix the bugs. Because eloquence itself can never be changed or updated anymore. When it comes to NVDA: Eloquence64 has much better indexing, but we only support the bundled DLL files. IBMTTS supports every eloquence version ever released, has parts written in rust, and less accurate indexing. But in exchange Chinese support works better, and it gets around a few Unicode and localization bugs.

@fastfinge OK that was very comprehensive and intriguing. Thanks.

@FreakyFwoof For my sins I maintain eloquence64, so I have to know more about this crap than I would really like to. The code for both addons is a terrible mess. But eloquence64 is…slightly better, mostly. Just because it’s trying to do far less. But when you have to work around libraries that haven’t been updated since 2003 or so, and can never be fixed, things are always going to be a mess.

@fastfinge Because people lock their source in a vault somewhere and don't let it out to play?

@FreakyFwoof @fastfinge More likely they dumped the drives in a landfill somewhere.

@FreakyFwoof Because Eloquence was originally written by Eloquent Technologies Incorporated, hence why it’s sometimes called ETI eloquence. But Eloquent Technologies licensed the code to IBM. Then the company was purchased by Speech WOrks International. Then Speech Works International was purchased by Nuance. Then Nuance was purchased by Microsoft, but the company that owns the right to Eloquence was split off into yet another company. During all that, the original creator of eloquence retired. So now it’s not even clear where the source code is, who (if anyone) still has it, or if they have it, who owns it. Because the compiled code was licensed and sublicensed and bought and sold. But did any of those transactions include the source code? Does anyone who still works at any of those companies that still exist even know? To make things even more confusing, you sometimes hear eloquence called ECI Eloquence, not ETI Eloquence. ECI is what IBM called the Eloquence Command Interface. So the dll files for ETI eloquence are often ECI.DLL.

@FreakyFwoof And for extra bonus fun, ETI Eloquence was created to commercialize the PHD work of Dr. Susan Herts. So she actually wrote the code at the two different universities she worked as a researcher for. The first thing that’s recognizably eloquence-like was written in 1984. Then she continued to evolve it, using her custom tools and compilers.

@fastfinge @FreakyFwoof Has anyone checked the warehouse from the end of Raiders?

@dhamlinmusic @FreakyFwoof I think the eloquence thing was probably the thing inside the arc that melted that guys face off when he looked at it.

@fastfinge @FreakyFwoof And to add to that, it's not even clear what source code even means in this instance. It is possible that somebody has the autogenerated cpp, but not the files that was generated from. Or that both exist and the compiler was lost. Or that people have the compiler but not its source. I find at least one of these to be likely considering that Apple compiles Eloquence for aarch64, but doesn't make any pronunciation changes.

@fastfinge @FreakyFwoof You just summarized like 90% of government computer systems…

@dhamlinmusic @FreakyFwoof Some of Dr. Susan’s research work was actually funded by The National Science Foundation and DARPA. So yes, the US government does have tentacles in eloquence.

@fastfinge @FreakyFwoof I mean when has something like this not been funded by DARPA…

@dhamlinmusic @FreakyFwoof That’s why the British Eloquence voice sounds so bad. It was, in fact, the US governments opening attack on Britain to provoke another war. Obviously.

@fastfinge @FreakyFwoof Assuming the Russian voice is actually mind control tech.

@dhamlinmusic @FreakyFwoof LOL you think it has a Russian voice? Nope! This was the 90’s. As far as developers were concerned, only the English alphabet existed. When support for Chinese, Japanese, and Korean was added, it worked by first converting everything into the English alphabet. Then reading it that way.

@fastfinge @FreakyFwoof Except for that one bug where there's one kanji character it can't pronounce because the low byte of it happens to be backtick.