Text to Speech

10 ^AM August 3, 2004

I've been researching Text to speech (TTS) software lately. There is a wide range available, some of quite good quality. I imagine that high quality TTS will become part of consumer operating systems sometime in the next ten years or so.

Meanwhile, there are plenty of web-based TTS software demos around. Here are my personal favourites. Enjoy!

Rhetorical - Accent: Focussed on the call centre market, with high-quality voices and excellent pronounciation. They do a credible Aussie female voice. Try this text with American and then Australian speakers:
I can say Sydney, Melbourne, Adelaide, Perth, Brisbane and Hobart, without sounding stupid.

Rhetorical also have "Valley Girl," the only TTS voice I know that can say "Like, y'know?" and "Well, duh" without any coaching.
Loquendo - Timbre: Loquendo's voices are full of timbre, like a television voiceover. I enjoy writing snippets from imaginary documentaries for their English male voice, Simon:
The Lesser Waddling Penguin can reportedly reach speeds of up to nineteen kilometres per hour. At such velocity however, even the slightest miscalculation can be fatal. Ooh! That has to hurt!
Cepstral - Sinister: Definitely at the cheap end of the TTS market, Cepstral would be a good choice for game developers. Try this: select the "Damien" voice, and the "Split Personality" effect, then get it to say:
All your base are belong to us!
AT&T Natural Voices - Realism: AT&T have spent years developing this product, and it shows. The US English voice, "Rich" would sound pleasant over the telephone. Try getting him to say:
Hello? Hello? I see. Someone will attend to it shortly. Please call again!

PS: Just wondering - has anybody out there used the Natural Voices SDK with the Java Speech API (JSAPI)?

By alang | # | Comments (4)

(Posted to Software Development and javablogs)

Comments

At 23:15, 03 Aug 2004 Pete wrote:

Alan, you never owned an Amiga, did you? Back in 1985 Amiga came with TTS in its first version. Admittedly it wasn't up to the standard of some of the above sites - but they don't show a whole lot of (significantly noticeable) progress for 19 years worth of work.

(#)

At 08:33, 22 Aug 2004 ron wrote:

my name is ron

(#)

At 02:24, 20 Dec 2004 atq wrote:

Thank you so much (to who ever gave the link to: Loquendo)!! I have been looking for a voice which is perfect for a documentry, Thanks to you , I now have some voices to use for my documentry!!

(#)

At 16:23, 26 Aug 2005 Comedy Portal wrote:

Easiest freeware text2speech technology I have found to uses is Microsoft's Agent Characters (see http://www.microsoft.com/msagent/ for full info). Most versions of Windows comes with these MSAgent characters already installed (aprox 75% of web surfers are using Windows operating systems) and waiting to pop up and convert text to read aloud speech. You can change the characters, pitch, speed and even animations of the characters with MSAgent... http://www.SirSeek.com is an excellent example of an online website that implements this MSAgent technology during searching, reading details of a site and best of all, the dictionary lookup allows you to have words phonically sounded out aloud (also check out their tools section for other unique MSAgent generators and web tools like Your IP Address, Talking Time/Clock, Stop Smoking Nagger that you set as homepage that gives random encouragement or health facts about cigarettes, etc). This technology can be programmed into software or web pages easily...

(#)