2018-08-02 19:36:57

Hi all,

I'm a bloody newbie in programming. I know this will take time but I think I could do it if I get help about it.

My dream would be to develop my own sapi 5 voice with my own voice that can be used in windows and screen readers as well. And I know there are two methods doing that, one of them is to read a lot of sentences to create the voice.

I really would be thankful if someone would take the time to explain it to me, perhaps someone is here who speaks german, german is my language. And by the way, the voice I'd like to create will be in german, too.

I really hope for positive answers.

Greetings,
Janagirl!

2018-08-02 20:31:55

Not gonna happen unless you commission someone to do it for you like one of the professionals. Yes yes there are ways like with that Circumreality thing and all, yeah that's really not going to provide a quality you're going to want. You are not going to do this at your home with any kind of good quality.

Facts with Tom MacDonald, Adam Calhoun, and Dax
End racism
End division
Become united

2018-08-03 04:47:04

Unfortunately, if you want to make a text-to-speech voice completely from scratch, you'd have to be familiar with high levels of linguistics, computer science, and maybe digital signal processing. Well, actually, this applies more to developing a text-to-speech engine. The important thing here is that there is a difference between a tts voice and a tts engine. The tts engine is the foundation upon which voices are built according to rules established by speech engine platforms like SAPI. Essentially, TTS voices fall into two categories: concatenation or formative. Concatenation  speech synthesis is what you're trying to achieve I think. In other words, you string together voice recordings such that speech is smoothly synthesized from a database of recorded phonemes. There are mechanisms by which this can be achieved through SAPI with maximum amounts of patience. However, the easier way would be to pay a company that specializes in synthesis algorithms to generate a voice for you. One such company is LyreBird. You can try contacting them to help you get started if this is something you really want to do. Keep in mind, however, that this will take a lot of commitment on your part.
Hope this helps.

2018-08-03 06:37:40

Probably a lot of money too, as you'd have to go into their studio to do it, not your home with whatever mic you could scrounge up and no type of recording environment suitable.

Facts with Tom MacDonald, Adam Calhoun, and Dax
End racism
End division
Become united

2018-08-03 19:06:10

I agree. To make a high-quality text to speech voice, you have to use a professional studio, which is expensive. I have that kind of stuff, but unfortunately, I don’t have currently the time to do that.

Feel free to contact me:
My facebook page
My skype

2018-08-04 08:49:41 (edited by defender 2018-08-05 08:43:16)

I've heard you can get buy quite well by simply lining the walls of a room in your house with that segmented foam board that you can slot together from a hardware store or some mattresses, installing quiet computer fans/using a fanless device like a tablet, mini PC or digital recorder, along with a decent shock mount, wind screen and XLR microphone, preferably into a mixer.
As long as you unplug all unnecessary noise making electronics and you aren't close to a busy road, do it at night ETC you should be fine.
And that's just one way of doing it as well, their are DIY home studio guides all over the internet.
Mind you that's still probably gonna be like a 600 dollar investment at least , but... I'm getting off topic anyway, because I don't know enough about the do it your self TTS programs such as Lyrebird to know if they can come through as long as you have a good recording set anyway.

2018-08-04 15:28:06

lyrebird can't be used with SAPI, it's cloud based, and machine learning based, and it has that weird metallic quality that is similar of hts based voices, such as RHVoice and the microsoft OneCore voices. I personally don't like that voice type because it doesn't perform at faster rates and consonants often kinda get lost in the fuzz. I'm not sure how well a German HTS voice would work. Plus lyrebird is commercializing to where you pay them and send them a voice model and they will create you a voice that is beyond the quality of what you can get with the now-named vocal avatar API. It's a contact process. But if you're gonna be able to make a home studio then you can probably do that, depends if they require offers from companies or if they'll do individuals. But like I said it's cloud based, unless they can compile the models to run offline.

----------
An anomaly in the matrix. An error in existence. A being who cannot get inside the goddamn box! A.K.A. Me.

2018-08-05 08:41:21

Ouch, I didn't know that...

2018-08-10 19:53:23

Post 7, thanks for clearing that up. I didn't know that LyreBird couldn't be used as a SAPI voice. At least if you know some basic HTTPRequest programming and such, you could easily integrate your own voice into AudioGames and other tts applications.

2018-08-10 19:57:03

hi janagirl
you can use m n l p
check off topic room guys, if you have link of this, poest it on there for me