rss
Update on NVDA and ETI-Eloquence
Due to the high demand to be able to use Nuance ETI-Eloquence or IBM ViaVoice TTS with NVDA, a few months ago, we asked for expressions of interest in purchasing Nuance ETI-Eloquence from NV Access. We did receive sufficient interest to make this viable and began negotiating with Nuance. More than three months ago, We raised several concerns with Nuance concerning the contract they provided. Unfortunately, Nuance have provided no response, despite several requests from us.
Given our concerns about the contract and Nuance's apparent lack of interest in our business, we decided to investigate IBM ViaVoice TTS again. This offers several advantages, including better pricing, more languages and, in our experience, far better customer service. The major obstacle we hit in our initial investigation was that it required installation due to its use of the machine registry. Recently, we devised and implemented a work around for this. We then contacted Wizzard Software to begin negotiations. Unfortunately, they informed us that they are discontinuing sale of the product, so this is no longer an option. We've contacted IBM in the hope that they may be able to help, but this is extremely unlikely.
We recognise the importance of ETI-Eloquence to many of our users. As such, we have expended significant effort and time attempting to provide a solution. Regretably, all our efforts have failed and as far as we know, there is nothing more we can do.
NVDA is developed by 
Comments
what about using maybe ivona text to speach, or even dectalk?
Dectalk seems unlikely, considering that even Fonix is no longer selling it and, even if you could convince them to do so, it would no longer be the Dectalk we all remember fondly. I wouldn't say Ivona is the best option either, simply because of its size and memory requirements. It wouldn't be very responsive except on high-end machines. It would probably be better, if anyone's considering the natural voice route, to go with Neospeech or Acapela. They're still relatively large, but even under SAPI 5 (which is laggy in and of itself), they're reasonably responsive. A direct NVDA driver for these would, I think, overcome the rest of the responsiveness issues. We really do need an official, built-in, alternative to Espeak. I've introduced several people to NVDA and their first reaction is, without fail, "That voice is horrible!" Most people don't realize that other voices can be installed, and if I hadn't shown them that, that would've been the end of their experience with NVDA. Eloquence is, unfortunately, a dead end. Most people don't like to hear this, but it's true. I personally don't even like Eloquence and have never understood most people's obsession with it, so I'm not all that saddened, but that doesn't change the fact that Espeak is just not good enough for those of us who have come to expect better tts. Espeak would've been amazing back in 1980. Now though, it just sounds awful and NVDA suffers from this since many people don't realize that the tts is separate from the screen reader.
An official built-in "natural voice" suitable for screen reading is currently not possible and probably never will be. This is almost always going to require an additional purchase or at least download by the user.
Hello,
After I developed the eSpeak's variants iven and iven2, which now commes at the pacage, I use other synth anymore. Previously, I didn't like eSpeak. Most brazilian and portuguese users have tested and they really likes them. They, like me, sas that iven and iven2 are really less robotic than robert and even max. I suggest you try.
Thanks
@jteh: I didn't mean built-in as in free, I just meant something official. I've no objections to paying for an alternative tts for NVDA if it has a direct NVDA driver and is officially supported. Built-in was, I suppose, the wrong word. @Ateu: I've heard Ivan and Ivan 2 and, at least for English, they're worse than the official variants. They've too much bass and every word sounds like someone trying to talk with their mouth open too wide. It may sound great for Brazillian Portuguese; I don't know as Portuguese is not among the languages I am familiar with. Each language, however, has different requirements for its pronunciation to sound anywhere close to natural, and neither Ivan nor Ivan2 work for English in the slightest, I'm sorry to say. For English languages, I think the Klatt variants sound the least robotic but the consonants are too sharp in comparison the the other sounds. It's too bad Svox Pico has so much static, as it's quite responsive and sounds rather pleasant otherwise. I wonder if a version of that could be created that isn't so optimized for tiny phone speakers and speaks clearly? Unfortunately, while I'd love to do this, I know nothing about how to actually generate/concatenate spoken phonetics into something coherent.