Simple question here. Where can I get the best, most authentic sounding voices for the avatars? I've purchased a couple of NextUp voice engines for a variety of uses but they still have an artificial sound to them. Any suggestions?

Hi jawalsh,

Welcome to the forum!

they all sound choppy to be honest - the best one's I've heard are from at&t -
http://www2.research.att.com/~ttsweb/tts/demo.php

when you use these voices in artificial intelligence software like http://zabaware.com the software itself adds a more life like cadence to the voice.

- the most diverse one's I heard are from http://www.voiceforge.com/demo?uservoice=Belle

and a few more like Cepstral, ivonna and neo-speech...

all 3rd party TTS developers ( less the microsoft royalty free that already come with windows ) want to charge you an arm and 3 legs for distribution because they are not royalty free - not even for non commercial youtube use.

if you don't mind getting a little technical you could always try your hand at Utauloids - which is a do it yourself synthetic voice software - never tried it myself but looks interesting - http://utau.wikia.com/wiki/UTAUloids

then there are voice changing tools - http://www.screamingbee.com/product/MorphVOX.aspx

Cheers!

• My Spotify  My Twitter  My Virtual Art Gallery
  
• My NFT    My Marketplace My Blog
  
• Music Sync Licensing

☯🐉

"To define Tao is to defile it" - Lao Tzu

I like Acapela: http://www.acapela-group.com/.

I use their online version, called acapela-box: http://www.acapela-box.com/, which allows me to use just the voices I need for as much as I need them (it is "pay-per-word"). The big advantage also is that the recorded voice files do not have the usage restrictions that come with most TTS packages (as discussed on this forum).

---

there are ways to make them sound more natural, but it might be more trouble than what it's worth. - it's the spacing and pitch that make it sound un natural.

gather a collection of common words from the service ...truncate them efficiently ...then add them to a sampler and play them from a keyboard, this would give you access to pitchbend, timing, timestretching etc. this way you buy a word once and build a library.

the musical equivalent of keyframing

• My Spotify  My Twitter  My Virtual Art Gallery
  
• My NFT    My Marketplace My Blog
  
• Music Sync Licensing

☯🐉

"To define Tao is to defile it" - Lao Tzu

I still dream to have some TTS SW that:

1. is able to assign different intonation, i.e. the possibility to assign different "moods" to each word or sentence (e.g happy, angry, surprise, sad, etc.) ...
2. is able to generate different accents (i.e italian ppl speaking english or vice verse, indian ppl speaking english, etc.)
3. has no legal issues based on where the final voice is being used
4. has a reasonable price

Why no SW house made this ? Is all of this more difficult (in terms of SW programming and algos) than building the whole iClone SW & tools ?
Maybe RL will think about this as the next "killing plug-in"...



My PC:
OS: Windows 10 Pro English 64-bit / CPU: Intel i7-9700 3.6GHz / MB: ASUS ROG Strix Z390  RAM: 32GB DDR4 2.6GHz / HD: 2TB+3TB  /  SSD: 2x512GB Samsung 860 EVO + 1x2TB Samsung
VB: Palit GTX2080 TI GamingPro 11GB / AB: embedded in the MB and VB (audio from the MOTU M4 I/F) / DirectX: 12

I use the MS SAPI5 ones as being redistributable usage of the voices would hold no legal ramifications as since you CAN redistribute the whole files etc for creating them in the first place.
You CAN edit the wav using other software too like Audacity using the Rovee plugin for variation of voices or use screaming bee on the wav file too.
None sound as good as using real people though.
I use Balabolka to export the wav files as iClone only sees MS Anna unfortunately.

"None sound as good as using real people though."

That's the point nr. 5, which I forgot in my list:

5. the voice shall sound real


My PC:
OS: Windows 10 Pro English 64-bit / CPU: Intel i7-9700 3.6GHz / MB: ASUS ROG Strix Z390  RAM: 32GB DDR4 2.6GHz / HD: 2TB+3TB  /  SSD: 2x512GB Samsung 860 EVO + 1x2TB Samsung
VB: Palit GTX2080 TI GamingPro 11GB / AB: embedded in the MB and VB (audio from the MOTU M4 I/F) / DirectX: 12

@swoop, yes, thats exactly how I feel about some 3D functions.

I could spend every waking moment insisting you learn about audio production in extreme detail and calling you names when you don't want to though, you just got to take it reaally slow, I'll learnz ya!!. :p

• My Spotify  My Twitter  My Virtual Art Gallery
  
• My NFT    My Marketplace My Blog
  
• Music Sync Licensing

☯🐉

"To define Tao is to defile it" - Lao Tzu

It's a challenge, just like animation... It would be easier to just have a bunch of real people talking, but not as much fun...

Apart from Pinhead, who sounds like Pinhead (Australian Lee), the voices in my Breezaway commercial were done with Acapela Box.

---

Hi Animagic,

nice... nice!! I like the tone of the voices on Acapela!
It is also easy to produce non-native English speakers by changing the text with the right phoneme (e.g. instead of Hi or I => type AI, otherwise an Italian/Spanish voice will pronounce it like you pronounce "E" in English) .

But.... why they do not show the prices and the characteristics of the SW ?
Also they mention about the NDA to be signed in order to get these "super-secret" information...
That's the most ridiculous thing I have seen about a SW house...

Honestly, they can only scare people away instead of understanding that they have a very good technology and just openly show what their SW can do and at what price.

My PC:
OS: Windows 10 Pro English 64-bit / CPU: Intel i7-9700 3.6GHz / MB: ASUS ROG Strix Z390  RAM: 32GB DDR4 2.6GHz / HD: 2TB+3TB  /  SSD: 2x512GB Samsung 860 EVO + 1x2TB Samsung
VB: Palit GTX2080 TI GamingPro 11GB / AB: embedded in the MB and VB (audio from the MOTU M4 I/F) / DirectX: 12