Shower thought: I wonder if any #blind kids today are growing up on espeak the way we grew up on eloquence? Is there a future where espeak is out of date and ancient, and people are clinging to it with the desperation people today cling to eloquence and dectalk? Honestly y'all, espeak read isn't that bad. I just wish espeak was usable on IOS; the available app is old and littered with bugs. #a11y#screenreader#accessibility
@fastfinge eSpeak will never be out-of-date and hard to run in the same way Eloquence is, because eSpeak is open-source and can continue to be ported forward, legally (unlike leaked DECtalk), by anyone who cares to.
@matt@fastfinge Mine is currently set to Mike. I haven't found a voice I can definitively say 'Yeah that's the one for me' but I'd like to do that, so that if I ever had to use it full-time, I had the thing configured just the way I need it to be.
@fastfinge@FreakyFwoof@matt I am trying the read voice now, and it is just ... ... ... ... I don't know, it jsut still has that ESpeak sound to it. Yuck!
@JamminJerry@fastfinge@FreakyFwoof@matt I can't use it, it's painful for me because of the recorded consonants. Their klatt modes still use the recorded consonants, which can be rather sharp depending on your listening device, and I'm from the US. I'd like my computer to speak in an accent of the country I'm from, but it's American English is still pretty sucky. There are at least custom voices to adjust the relative formants and all that to craft the perfect klatt voice, and you can turn down the recorded consonants, but not off. I don't like some of those new ones like edward2, they sound like they're ringing and I can hear the upper formants as distinct pitches.
@JamminJerry@FreakyFwoof@matt@x0 If they didn't have so many issues with nonstandard words, I'd like the blastbay tts voices honestly. But they're going to require either fixes on Philip's end, or absolutely massive custom dictionaries.
@fastfinge@JamminJerry@FreakyFwoof@matt@x0 Also it has really intrusive english in other languages, spanish for example. If you have words ended in "al", "ción" or "able", One Core will Englify those words. Gooogle tends to do the same in Spanish US but it uses English phonemes, One Core just use Spanish ones and it sounds terrible.
@FreakyFwoof@JamminJerry@fastfinge@matt Mark is the only decent American voice on that, can get kinda fast, but nothing like the one person I know who uses eSpeak at 600 WPM. I also don't like the characteristic blurry sound of that synthesis method, and mark has some, strange pronunciation.
@FreakyFwoof@x0@fastfinge@matt I am not a fan of one-core, but if I had only it, and ESpeak on my system, and I had to choce between them, I would pick one-core without hezitation. I can't remember which voice it is I kind of like. Let me see here. OK, the only voice I even kind of like is mark.
@dhamlinmusic Two? I knew about the neural voices add-on, which bundles the data itself and uses the DLLs via an IPC process, what else was there? THe only one I knew about was natural voice SAPI adapter but that's not specific to NVDA. @fastfinge@JamminJerry@FreakyFwoof@matt
@matt Yes, "anyone who cares to" is the key phrase here. From what I understand, espeak's code is terrible because it was written for Risc OS. And the overlap between blind people who know C++ and blind people who want a speech synthesiser kept up to date is going to be vanishingly small. The Espeak IOS app is a perfect example. It hasn't been updated in three years, even though everyone who tries it is aware of the bugs, and many people want it fixed, and it's not really useful in its current state because of the constant crackling. Just because something is open source doesn't mean it will be maintained.
@matt Heck, even the Espeak on the Google play store is only updated once every year or two. Windows and Linux are really the only platforms that are maintained. If whoever's in charge of Linux builds stepped away, I could see the updates stopping there, too. Windows is only safe because NVDA depends on it for now. But even they switched to One Core by default. So I can absolutely imagine a day where espeak is just as hard to run as eloquence or dectalk. It would be more legal, sure. But that doesn't always mean easier.
@alexhall@fastfinge@matt I would honestly even say Linux and maybe also Android is getting the most love and the library. From a Windows/NVDA context, yes, it's getting updated. You want a SAPI version that'll work with other screen readers or apps though? The only version that still works is an ancient version pre EspeakNG which is also 32-bit only. EspeakNG never had a working SAPI version that I'm aware of - there were some builds that would show up in SAPi settings but not actually function.
@fastfinge Very much so. Had I used NVDA first instead of jaws, I would’ve latched onto too eSpeak. I still very much like it and would probably use it with JAWS if JAWS' SAPI implementation wasn't... bad.
@fastfinge Michel voice variant for me. Or Michael, no idea why they both exist since they are identical, or appear to be.
That said, what about the bugs in the espeak app on ios? It works pretty decently for me here at least, the crackling issue only happens after past a certain speech rate and I've never had cause to go that far with it.
I know there is an issue on gh about updating it and the dev said it would be done, no idea where that went off to if anywhere.
@fastfinge Oof yeah in that case, I get you, definitely.
Well, in a way I'm still glad we have it on ios, because on android it's, excuse my french, shit. Or was, last time I had android in hand. No way to set the variant, the app wasn't even belonging to espeak on the play store, it was full of ads, and android 10 broke it so bad it became unusable.
Meanwhile ios side, espeak survived 16, 17, 18 and 26 and has no ads, you get access to all variants with a pleasant enough interface.
Kind of crazy because eloquence for me used to be my go to. But seeing as ibm tts is shit on linux, that I use espeakup anyway, well. I tolerate eloquence mostly as a fallback, these days, whereas before I absolutely wanted nothing to do with espeak, until I found out about Michel variant.
@xogium Okay so, for gits and shiggles I looked closer into the state of espeak on Android. The app in the play store is broken. However, you can download an APK from the official github, and that works fine; it's up to date, modern, and has no ads. Why isn't the official APK from the espeak-ng github on the play store? Who knows! It's open source; you'll take your inexplicably broken things and like it that way, darn it! Oh, maybe it's not on the Play Store because it's on F-Droid? Hahaha no! Get your APK from github just like you do for Windows apps; no package management, for you, buddy.
@miki@fastfinge Agreed. eSpeak is not only available in many languages but it's open source, meaning that it's currently free of cost to a variety of users and this will likely ensure continued development, at least from a core group of developers. However, I think that many people in some countries will consider it to be not as useful as the commercial, more human-sounding TTS engines.
@DavidGoldfield@fastfinge More importantly, it used to be the default speech synthesizer for NVDA users (which is the primary screen reader for the "rest of the world")
