AI Zone Admin Forum Add your forum

NEWS: Chatbots.org survey on 3000 US and UK consumers shows it is time for chatbot integration in customer service!read more..

RSS AI Zone  Forum Home  >  Development  >  Tools  >  Thread
TTS playing pre-recorded mp3 and not single letters synthesis
  
 
Posted: Sep 13, 2015

Many TTS soond incredibly robotis, or all, comprehensive of SIRI and GOOGLE VOICE.
I think that if a TTS has the fortune to focus around an argument, thebest would be to record some hundred words, with perfect sound, instead to synthesize the words.
A good example are many navigation systems that not synthesize the words from the letters, but they have pre recorded the full words, and they can do it easily because they use few words: TURN LEFT, TURN RIGHT and few more.
So.. where I cna find a TTS that permit me to add my custom words instead to synthesize the words from every single letter?
 

 
Posted: Sep 13, 2015
   [ # 1 ]

I use https://www.cereproc.com/ which I have always found very realistic. I am from the north of England and their Jess voice is just like talking to someone from my area.
 

 
Posted: Sep 13, 2015
   [ # 2 ]

Can’t you program your own? I once programmed one in a day: Split a text at spaces, play wave files named after each word, and monitor the output channels to tell when sound stops and the next file should play.
However, a hundred words weren’t nearly enough (an average human vocabulary has 8000 words), many words were missing, and each sound file needed many edits to match up with consequtive words. The problem with this approach is intonation: When you use prerecorded word sounds, they’re all spoken out of context and you. gEt. A. soundlike. thIS-. Which is why they developed dynamic voice synthesis.
 

 
Posted: Sep 13, 2015
   [ # 3 ]

I know that it should be quite easy, but I am not a developer..
 

 
  login or register to react
‹‹ AI + internet = ?      Defining own tags in Treetagger? ››