Realistic voice for Text-to-Speech
I think you mean Text-To-Speech (TTS) since you are talking about a robotic voice. TTS engines take written text and voice it back, whereas speech recognition engines understand human speech and convert it into a machine readable format. Loquendo has the most natural sounding TTS engine I have heard. They have a version for Linux. They have an interactive demo you can play with to hear how great it sounds. There TTS engine can take special characters in the text to do things like provide emphasis on phrases or even make the voice laugh or cry. Not many TTS engines do this.
After weeks of researching the same question I found the voices from Ivona (here) and Loquendo (here and here) to be the best TTS voices available for Windows. Only Ivona lists prices on their website though. To actually use the TTS voices on your Windows PC I recommend Balabolka (free), Ivona MiniReader (free) or Ivona Reader.
Next check out Clearly from Evernote. It is a browser extension currently only available for Google Chrome. It provides TTS for premium customers only. The (iSpeech powered) TTS voice is not as good as Ivona's or Loquendo's but it gives you a nice reader view on blog posts/articles by stripping away distracting page elements. You can also set it to auto-highlight the currently read word!
Check out ReadSpeaker too which can be implemented by website owners into their site. Readspeaker provides auto-highlighting of the currently read paragraph/word, auto-scrolling and the ability to change the TTS speed.
[Edit:] There are some free useful Google Chrome extensions that are powered by iSpeech as well.
You can buy a ready made device . http://www.textspeak.com makes human sounding text to speech products, boards and boxes. They have 20 languages and simple are the best sounding TTS you can find.
There is no license or development...just buy the box if you only need one voice output.
(This is NOT a SW solution, but we used it in a paging systems... 5 minutes to install)
From their site
"TextSpeak Embedded Text-To-Speech modules series convert ASCII text to a natural, clear voice with unlimited vocabulary. The small footprint, plug-in solution accepts wide range of input data to generate real-time speech for Security Transit Medical and Industry.
http://www.textspeak.com/oemtts.htm