Hey gang, got a quick question about Text To Speech technologies. Does anyone use them?
Hey gang, got a quick question about Text To Speech technologies. Does anyone use them? I'm thinking of having one for station announcements and positioning, but am on a limited budget. I do use VOA news and NOAA MP3 stream for weather. Is there good software that's either free or dirt cheap? I realize I might get subpar performance, but if its recognizable (hint: NOT NOAA's old "Perfect Paul"), I'd love to look into it. Thanks!!!
For free, you can enter short phrases into these three web pages and select various voices.
http://www.neospeech.com/demo/demo_text.php
http://www.research.att.com/~ttsweb/tts/demo.php
https://www.cepstral.com/demos/
For inexpensive software, TextAloud is pretty good: http://www.nextup.com/.
Nextup also sells several voices. However, if you want NeoSpeech Kate and Paul, you can get them cheaper here: http://www.smarthome.com/12662.html
I use a synthesised voice for time announcements in ZaraRadio, but personally, I wouldn't use it for regular announcements.
Edit: You can also get TTS voices from Loquendo, Scansoft (or whatever they are calling themselves this week) and others.
In my opinion, the AT&T and NeoSpeech voices are the best sounding of any I have heard. The NeoSpeech voices sound more natural, but the AT&T voices are slightly better with some pronounciations.
For a free program that is similar to TextAloud, DSpeech is pretty good. I think TextAlound does a better job, though.
Currently I use a free/Open Source package for Linux called Festival to automatically read dynamic content from RSS feeds and other sources for daily news, weather, and other purposes. Out of the box it's not the best sounding but voices can be modified or designed from scratch if you like to tinker. I'm still working on getting the extended tools working so I can get the hot sounding Swedish female voice to read slashdot and AP news ;). There's another package called Flite that is a simplified version of Festival that is cake to install (It was designed to run real-time menus for PBX's), but it doesn't sound quite as nice.
Festival: http://www.cstr.ed.ac.uk/projects/festival/
Flite: http://www.speech.cs.cmu.edu/flite/doc/flite_toc.html
The biggest issue is to do pre-processing on your input text before feeding it into the t2s software. I've got a sed script that de-funkifies most text scraped off news portals as well as translating some abbreviations and acronyms to their full proper spoken phrases that I would be willing to share. You'll have to do your own screen-scraping though, I'm not interested in giving away my automated station DJ just yet. If you know any Unix scripting languages (korn shell, perl, php, DOS batch, ect.) it's cake to whip up some cool sounding reports at regular intervals for your station.
dosman [@] packetsniffers.org
I also use and can recommend;
http://www.research.att.com/~ttsweb/tts/demo.php
With some experimentation with phoenetic spelling and different punctuation I can usually get reasonably good results.
Good luck!
-Dave
Does anyone know how to use Dspeech as a command line - supposedly it can be done (but maybe with an earlier version)? Does anyone know of any other TTS software (free or inexpensive) that will act in command line mode. The only other one that I've found that will obviously do that is espeak, and it sounds pretty awful (only uses the SAPI4 engine, not SAPI5).
"Read text aloud with support for SAPI4 or SAPI5 compliant voices."
http://www.mycmd.com/multimedia/convert-text-to-speech
Also check assistive technology web sites.
Experimental broadcasting for a better tomorrow!
Well, since I posted my question about Dspeech, I've found one program that does support being driven by command line - TextToSpeech by Alive Media. It's not free ($29.95) but appears to do the job (TextAloud by Nextup wanted an extra $250 to add this functionality to their similar graphic interface). In the tests I've done, the NeoSpeech Kate voice sounds the best for my application (ferry updates for the little island I have my station on). I'd still like to find out about Dspeech, though.
Overall, I'd have to say that I'm pretty disappointed in the quality of most of the software and voices that I've tried - they certainly don't live up to expectations (even with the pricier ones). Perhaps if I had thousands of dollars to spend...oh well.
Not practical for real time stuff, but if you want to do some pre-recorded stuff, changing the text to meet the software's ability can help.
Experimental broadcasting for a better tomorrow!