Speech synthesis/TTS module that DOESN'T sound like a Speak-N-Spell?

Projects General Guidance
1

Doing some research for a new project which requires decent text-to-speech capabilities for a vision impaired friend. I'm entirely unimpressed with the SpeakJet's output, which is jarring and quite artificial sounding. Parallax used to sell a serial module which sounded OK, but it has been discontinued. TextSpeak (TextSpeak | Board and Modules - Text To Speech Embedded Audio, voice synthesis, chips, Ethernet, Wireless and RS-232 audio players.) sells a beautiful-sounding TTS module but it is prohibitively expensive for this project at $300USD. I'd like to keep the module price at or below $100USD.

Does anyone have any experience in this area? I really don't want to resort to a PC-based solution, though the Festival project has some pretty decent sounding voices. If looking at arduino solutions am I better off building a discrete speech system (eg, piecing together phrases from discrete words)? Any input here would be appreciated.

2

buzzdavidson:
Doing some research for a new project which requires decent text-to-speech capabilities for a vision impaired friend. I'm entirely unimpressed with the SpeakJet's output, which is jarring and quite artificial sounding. Parallax used to sell a serial module which sounded OK, but it has been discontinued. TextSpeak (TextSpeak | Board and Modules - Text To Speech Embedded Audio, voice synthesis, chips, Ethernet, Wireless and RS-232 audio players.) sells a beautiful-sounding TTS module but it is prohibitively expensive for this project at $300USD. I'd like to keep the module price at or below $100USD.

Does anyone have any experience in this area? I really don't want to resort to a PC-based solution, though the Festival project has some pretty decent sounding voices. If looking at arduino solutions am I better off building a discrete speech system (eg, piecing together phrases from discrete words)? Any input here would be appreciated.

I'd go with Festival on a BeagleBoard (kinda PC - think of it as "embedded PC"); the problem with the more natural sounding solutions is the higher amount of processing and memory needed needed...

3
4

that video creeps me out. seriously. one ugly robot-thingy.

5

If you can, see if you can track down a Digital Equipment Corporation (DEC) DecTalk module. Serial interface, multiple voices, still used by Ma Bell in some places.

If you can get away with it, maybe you might consider an MP3 shield, if it can switch words/phonemes fast enough.. you can then just record in whatever voice you would like...

6

@focalist: thanks, I'll look into the DecTalk.

That video is way creepy. Interesting tech though!