Third-party speech synthesizers coming to macOS, iOS, and tvOS.

By Zachary, 20 August, 2022

Forum
iOS and iPadOS

Hi all,
I'm sure some of you may have already seen this announcement on Twitter, but Apple is rolling out an API for third-party speech synthesizers to be used with accessibility features on their platforms. Known as the Speech Synthesis Provider Audio Unit, this should be coming out in the initial releases of iOS 16, Mac OS 13, and tvOS 16. Hopefully companies will get on board pretty soon. Here's a link to the developer documentation for anyone wanting to take a look, and please let me know your thoughts and which voices you would like to see. https://developer.apple.com/documentation/avfaudio/avspeechsynthesisprovideraudiounit

Options

Comments

By kool_turk on Wednesday, September 7, 2022 - 22:05

As good as this sounds, until we actually see results, there's no point getting excited.

Until we actually see 3rd party synthesizers to download, then, and only then would be the time to get excited.

You know when Eloquence came out of nowhere? That's the sort of thing I like to see.

None of this hyping up business.

By Igna Triay on Wednesday, September 7, 2022 - 22:05

While this is great news, I'm with @2, I don't want to get my hopes up until I see the results and this happen.

By KE8UPE on Wednesday, September 7, 2022 - 22:05

Hi,
I heard that you can use this to create a tts version of your own voice.
If i'm able to figure out how it works, I plan to make one.
I've always been curious what mine would sound like.

By Ekaj on Wednesday, September 7, 2022 - 22:05

This is very cool and I hope something comes of it. I read somewhere--think it was Twitter--that Cereproc has a Donald Trump voice. Lol I'm not saying what political party I'm in, but this would be great to have. For now though, I'm excited about the voices that are actually coming in iOS 16 and Mac OS Ventura. Having said that I'm with those who say don't get your hopes up, at least for the time being. Apple's got a lot on the plate as it is.

By Trenton Matthews on Wednesday, September 7, 2022 - 22:05

Considering you can get Cereproc voices for only 0.99 USD on Google Play:
https://www.cereproc.com/en/android_tts_voices
, and yes they do have a Donald Trump TTS voice, demoed on Cereproc's own YouTube channel:
https://www.youtube.com/watch?v=Vlt57PVTzNQ
, I'd love for that voice company to show up on IOS devices! Especially since ya can already get them on Mac OS, and Windows of course, though the Mac/Win versions are 29.57 US Dollars.

By Exodia on Wednesday, September 7, 2022 - 22:05

This is very interesting, however, I am with the others. As exciting as it is, I don’t want to get excited. However, I would love to see some really old voices not just for Mac, but other synthesizers as well. And the reason I say that is simply for nostalgia. Soft voice would be one of the Synthesizers I would love to see. Some of the voices from that could be just under the novelty section. Such as colossus, fly, munchkin and all that other sort of stuff. I would also love to see dectalk added to it. I know a lot of people these days would prefer synthesizers with mostly natural sounding voices, but I kind of like some of the other robotic ones for nostalgic purposes.

By Amir on Wednesday, September 7, 2022 - 22:05

If VoiceOver can use Espeak via this approach, a very good number of languages will be supported. Moreover, as natural as the new Persian Dariush voice in iOS 16 sounds, it has awful intonation and pronunciation issues though they are not related to Apple. Espeak, as robotic and mechanical as it sounds, supports languages like Persian more satisfactorily.

By techluver on Wednesday, September 7, 2022 - 22:05

I mean... if this really works, there is absolutely nothing to stop people from porting eSpeak. Speak is open source so it is super doable.

By emassey on Wednesday, September 7, 2022 - 22:05

This almost certainly means that eSpeak will be available on Apple platforms, which is great news! Since eSpeak is open source, even if the eSpeak developers do not choose to make an iOS version of it with this new API, anyone who wants to and has the required skills could do it. If the espeak-ng fork is used and the Sonic audio library is included, there is no maximum limit for speech rate, and I have been able to understand it at 900 words per minute, so it would be a great synthasizer for Voice Dream Reader or other apps that read eBooks using text to speech.

By techluver on Wednesday, September 7, 2022 - 22:05

Quite honestly at that point it may be feasible to collect money from the community to hire a fiver developer to do this haha.

By emassey on Wednesday, September 7, 2022 - 22:05

Here is a GitHub issue someone created in the espeak-ng repository a few days ago about this. No one has responded yet but hopefully that should help to start the process.

By Manuel on Wednesday, September 7, 2022 - 22:05

Are you sure that this new API is indeed for providing speech synthesizers? I cannot find any reference in the developer docs that implies that.

By Zachary on Wednesday, September 7, 2022 - 22:05

Yes, this API is intended for developers to create new speech synthesizers. This is distinct from AV Speech Synthesizer, which from what I understand is for use with apps that utilize TTS to read content. Speech Synthesis Provider is providing synthesizers to the system for use with AV Speech Synthesizer.
This is based on audio unit extensions, which were originally designed to be used with audio software such as logic pro and GarageBand to add third-party effects and instruments to these applications, but make them available systemwide so that the same plug-ins could be used across different software. This is a similar concept, except the audio unit and question is a speech synthesizer providing voices to the OS, and intern to all apps using the speech synthesis framework, such as VoiceOver and Speak Selection.
The guy that originally tweeted about this was involved in its development when working at Apple, and has confirmed that this is the intended use for this new API. For anyone who is interested, there is a template in the latest beta of Xcode that will allow you to get started with Speech Synthesis Provider. I would encourage anyone who is able to take a look and maybe attempt to port something open source to make sure this is working as intended, as I'm sure there are problems that need to be corrected.
I have a feeling that Eloquence was implemented using this API, although of course I have no concrete data on this.
Please let me know if you have any further questions.

By ming on Wednesday, September 7, 2022 - 22:05

I would like to see if it can have all the microsoft voices or Neo speech voices also the accapela voices as well.

By sockhopsinger on Wednesday, September 7, 2022 - 22:05

I'd be very happy with a Katie Morgan or Natalie Portman voice. :)

However, since neither of those are likely possibilities (a bloke can dream), I'd love for the Neospeech voices to port over. I don't think they get enough credit for being excellent quality.

By Knut on Wednesday, September 7, 2022 - 22:05

How would these new third-party voices be distributed? Through the App store? Furthermore, the possibility of creating new speech syntheses based on your own voice should be accessible via the user interface: just as you train your iPhone/iPad/Mac to recognize the voice for Hey siri, why Apple doesn't create a user interface in which VoiceOver guides you in reading sentences, thus generating a neural voice based on your voice?

By techluver on Wednesday, September 7, 2022 - 22:05

First, I am glad for that clarification. I spent 5 minutes looking at the devs docs but didn't really take much in lol.
second, maybe that's the next step, being able to make your own voice. :) Personally I can't see myself doing it but...

By The Tetris monster on Wednesday, September 7, 2022 - 22:05

Hi guys. This is absolutely fantastic news. Maybe this will finally allow us to get an Afrikaans voice. As long as it’s not espeak, I’m good.

Not to say that the entire engine is bad, but I find the Afrikaans voice very difficult to listen to for some reason. I’ve always wanted Apple to make an Afrikaans voice for Siri, as well as give us The ability to use it with Siri and dictation. I’d love to see an Afrikaans version of eloquence, but I know that’ll never happen LOL. As for what voices I would like to see, I still want the female version of Alex, maybe different variants thereof such as teenager adult and senior citizen, and I’m not just talking about differences in pitch.

By techluver on Wednesday, September 7, 2022 - 22:05

Honestly your best bet for an afrikaans voice would be to talk to the qfrency people. Not that that voice is particularly amazing but it's better than eSpeak haha.

By Ekaj on Wednesday, September 7, 2022 - 22:05

@Exodia: I vaguely remember Softvoice. Never used it but Patrick Perdue had his Drunk Willowpond. Don't know if that's still available.

By John Gurd on Wednesday, September 7, 2022 - 22:05

Rather than new voices, nice though this is, i'd rather there was a way to fix the bugs in Voice Over, or for 3rd parties to write new screen readers which can make a better job of accessing the internet, etc, and can be updated with useful and time-saving features.

By Abdulaziz Faqihi on Wednesday, September 7, 2022 - 22:05

Since it is now possible to use third-party synthesizers, I hope that Acapela voices will be available on IOS

By Pax on Wednesday, September 7, 2022 - 22:05

why you prefer talkback, when it's laggy and lacking basic capabilities?

By GeorgeStark on Friday, October 7, 2022 - 22:05

Definitely i'd wish for eSpeak to be added into iOs, since it supports many languages, including my own. It's barrier which's preventing me to use iOs device, rather because of it i'm using android.

By kool_turk on Friday, October 7, 2022 - 22:05

I think you would download them through the app store, at least, I think that's how it's done on Android.

You download that company's app and use that app to get the voices.

But Apple doesn't have app stores with in an app store, at least, not really, so not sure how something like that would work.

I know Choice of Games was able to pull it off but they were kind of forced into it because they have like a tun of games.

We'll just have to wait and see.

By Piotr Machacz on Friday, October 7, 2022 - 22:05

Yep. THe voices are a kind of plugin, and Apple requires plugins to be part of an app. So if a synthesizer were to be released for iOS or Mac OS for that matter, it'd just be an app that you need to download from the store. Then while most of the functionality would be in the TTS plugin itself Apple requires that the app does at least something, which could be as simple as directing you to settings to change the voice or have more functionality like downloading and purchasing voices through in-app purchases, a separate dictionary and so on.

By KE7ZUM on Friday, October 7, 2022 - 22:05

that cannot happen due to some api issues, mocrosoft have indicated that this will not happen for the foreseeable future.

By Vsevolod Popov on Friday, October 7, 2022 - 22:05

I really doubt that third-party companies will really develop new speech synthesisers to use with iOS and MacOS.
I don't know why.
As someone has already stated, I would prefer Apple to fix bugs that aren't fixing for 6 years with Russian voices and capital letters, and yes, I do report bugs.
There are even more bugs for multilingual users, but Apple prefers to include Eloquence instead of fixing old bugs with braille screen input and "unpronounceable" message when entering text in languages that are not based on Latin letters and many more bugs.
Even if companies create new voices, I would prefer to use Acapela group's voice instead of Cerence, I am really tired of iPhone's voices. RHVoice, eSpeak. But I really doubt that there will be someone who will maintain eSpeak, it will be needed to update it every commit to stay with the latest version.

By Vsevolod Popov on Friday, October 7, 2022 - 22:05

Well, I have an iPhone 13 and a device Samsung Galaxy S10 with Android 12 and I can tell that Talkback is much more sluggish than VoiceOver. It is changed in Android 13, Talkback gestures will be processed through the system itself, but I feel like I won't be able to test and tell if it's true or not.
Also, there is a fantastic image recognition feature in VoiceOver that I don't know if ever exists in Talkback.
If I still decide finally to switch to Android due to a situation with removing necessary apps from Appstore of my country, I will regret missing this image recognition feature that in iOS gets better and better.

By Labsii on Friday, October 7, 2022 - 22:05

While this appears to be likely, this is not documented in any way. There is no clear wording from Apple that this is currently possible, let alone examples and templates.

My opinion is that this was released as new voices in iOS 16 were developed using this and needed the backend to be there.

However it is likely that Apple hasn’t really intended this to be used by 3rd parties yet. And while technically it can probably work even now, making something like that work without support from Apple is a task that is enormous.

So probably you will need to wait for some new releases from Apple that would include a complete toolset and documentation in order to expect some 3rd party voices.

By Lily Rose on Monday, November 7, 2022 - 22:05

I know this is kinda old but yeah I think Acapela voices on IOS might be nice. Going to have the same voice I use for my brailleNote, or used to until I got it replaced. Yes now I use Sharon, and I honestly think it (as the default voice on BrailleNote Touch +) is way better than Samantha. My favorite voice is Tracy, cranked up nice and fast, and so high-pitched and squeaky it sounds like a kid. LOL I really wish the kid voices were available on the BrailleNote Touch. But that's beside the point. I'm just saying acapela voices would be awesome.

By ming on Monday, November 7, 2022 - 22:05

anyway...
hope we can have some voices later

By Levi Gobin on Wednesday, December 7, 2022 - 22:05

Well, our first one is here, eSpeak. I myself prefer eloquence over eSpeak. The ocopella voices would be nice. Put on heather or Sharon and read or do some work? hell yeah.
I do agree that apple needs to fix VoiceOver bugs. I myself don't know about much of the other voices that aren't English, but if they don't work well, apple needs to fix them.
(apple accessibility team, I'm looking right at you! get on it!).
What speech synths do you think we will sea next?
I really hope that they add realspeak solo direct, but that will probably, if not definitely never ever ever happen... I just like the real speak voices much better than vocalizer, especially the Jill voice.

By Levi Gobin on Saturday, January 7, 2023 - 22:05

As you all know espeak has been out on the iPhone for several months now. I’m starting to get tired of vocalizer voices, Siri voices, and eloquence on the iPhone. Do you guys know of what other options I have as of now? If I have to keep using eloquence I can, but I would like something better.

By roman on Saturday, January 7, 2023 - 22:05

Hey.

I do not like any vocalizer voices because they have a unnecessarily weird inflection.

Alex on the other hand, is way smoother than the rest of the vocalizer voices.

I really hope that apple can work on eloquence voices because vocalizer is sucks.

cheers.

By ming on Saturday, January 7, 2023 - 22:05

the vocalizer voices in IOS 16 i not good.
especially using Chinese languages

By Ekaj on Saturday, January 7, 2023 - 22:05

Apologies in advance if this is a duplicate comment. But I'm really looking forward to using eSpeak on my current phone or the next one, and on the Mac once that is possible. I used eSpeak with NVDA, and really liked it. I can't wait for other TTS engines such as those celebrity voices!

By Matthew Whitaker on Saturday, January 7, 2023 - 22:05

Hey, it would be cool if someone could get TrueVoice to work. Microsoft SAPI 4 and SAPI 5 voices would be cool too. I"m sure it is pretty easy to do so.

By Levi Gobin on Saturday, January 7, 2023 - 22:05

After googling for like ever, I couldn’t find anything else. That’s why I posted on these forms to see if you guys knew of anything else.

By fatih on Saturday, January 7, 2023 - 22:05

Espeak is good, although we cannot use some languages that the iPhone supports as default languages on voiceover, for example Persian and Kazakh.
When you set the iPhone language to Kazakh for example and you set up the espeak Kazakh voice from the espeak-ng app in the voiceover languages section, voiceover still uses and recognises Russian as the default rather than switching to Kazakh.
The same goes for persian as well: Even though vocalizer voices has native support for Persian voice, when you set it up and try to read a Persian text on a webpage in Safari, Voiceover still continues to use the default language rather than switching to Persian.
I use my iPhone in Russian though and I'm happy with espeak. Thanks to the developer Yuriy Popov, who is from Russia.
I think it'd be good for people who use espeak to rate it on the app store and write some comments. It'd increase the app's visibility and would encurrage the developer to further improve espeak on IOS.

By ming on Saturday, January 7, 2023 - 22:05

anything will be released in the near future ?

By The Toy Fanatic on Thursday, September 7, 2023 - 22:05

Update,rh voice is now available, but it does not support English as of right now, if any app developers are reading this, I would love to see Ivona voices come to iOS, if you can, and if you have the source code, please make that happen