github.com/tgeczy/Brailab-wrap…
Hey @erion, do you know if it's safe to share the tts.dll (compiled of course) with a release of the driver in the wrapper repo, or licensing prevents this?
GitHub - tgeczy/Brailab-wrapper: A DLL wrapper for the Hungarian Brailab Speech synthesizer.
A DLL wrapper for the Hungarian Brailab Speech synthesizer. - tgeczy/Brailab-wrapperGitHub
Erion
in reply to Tamas G • • •Tamas G
in reply to Erion • • •Erion
in reply to Tamas G • • •Yes, that's what I mean as well. NetMedia (through us) made a contract with the makers of Brailab to distribute it for Hungarian NVDA users in binary form, along with a custom addon. I've created the initial version of the addon, this was before audio device support was implemented for NVDA.
The addon itself and the Brailab library was very likely leaked during the internal testing process.
Around this time, Vocalizer for NVDA was released, which provided more human-like voices. Despite the fact that we couldn't fix a lot of the dictionary entries incorrectly defined by Nuance, NetMedia started distributing it, scraping the Brailab addon entirely. So technically, NetMedia has the distribution rights.
Tamas G
in reply to Erion • • •Erion
in reply to Tamas G • • •You should, if you ask me. I totally agree.
This is kind of a grey area, because the addon would have been free, so it's up to you.
BMEITT is actually working on AI versions of Profivox, but of course they don't make it available to those who would really need it and when they do, they make exclusive deals, see Jaws. Trust me, we have tried far beyond what was possible to implement a good Hungarian synth for NVDA, even going as far as doing our own, partnering with Cereproc and Meridian One, constant refusals from BMEITT, and of course a lot of business dinners, etc. The state of Hungarian TTS is quite bad right now.
Tamas G
in reply to Erion • • •I installed Hungarian JAWS the other day. I didn't see Profivox listed, but I could manually add it. Looks like they default it to Vocalizer Mariska, which is horrible. Am I wrong in this? I don't understand how Hungarian blind people are putting up with this day to day if that's true, even old problems with things like double-S are still in it.
Erion
in reply to Tamas G • • •You tell me, at least on Windows there are choices, Apple makes you suffer lol. I can't even list what's wrong with it honestly, there are just countless issues popping up every day.
I think the price of making a good synth wouldn't be as huge today, at least if you create a model. Training costs and around 40 hours of data should produce reasonably good results with Piper for example. Maybe around 10K or less would cover it.
Tamas G
in reply to Erion • • •Erion
in reply to Tamas G • • •Tamas G
in reply to Erion • • •Erion
in reply to Tamas G • • •Ah I didn't know the Axelero connection, but I knew about Speech Tech owning it. I guess Axelero supported them financially. Later on they sold SpeakBoard with the same voices.
I doubt any SDK for it exists today, I'm not sure they were open to third-party licensing back then.
Tamas G
in reply to Erion • • •Erion
in reply to Tamas G • • •Tamas G
in reply to Erion • • •Winter blue tardis
in reply to Tamas G • • •Tamas G
in reply to Winter blue tardis • • •Winter blue tardis
in reply to Tamas G • • •Winter blue tardis
in reply to Tamas G • • •Erion
in reply to Tamas G • • •Winter blue tardis
in reply to Erion • • •Tamas G
in reply to Erion • • •Erion
in reply to Tamas G • • •