I have been trying to find a good solution to TTS on Linux for about 8 months now and have done quite a bit of reading around. I have decided to post information on what I have found already as it may be useful to someone out there… However the question I’m trying to answer really is how to get a high quality voice like the ones from Cepstral working in Ubuntu or any other linux distribution? www.cepstral.com a closed source program costing around $40. I can get it running fine in the command line but can’t find a suitable interface using it on a day to day bases.
I require a tts program to help me proofread nearly everything I write as without it I almost always have to many mistakes. There is not only good tts for Windows and Mac but for Android and ISO, so it should not be beyond the realms of possibility to have a working solution on linux.
First for the Linux one tts, festival, its very impressive that all this work has been done both free and open source. I have tried out some of the different voices, the ones people say are the best but they are still not good enough. Someone also wrote a nice simple bash script to use with festival and xsel. Then you can set a keyboard command to trigger it you have a very nice tts system. The script is from: http://hak5.org/episodes/haktip-51 & http://www.ghacks.net/2010/10/09/lin...with-festival/
Code:
#!/bin/bash
xsel | festival –tts –pipe
Now if we could find a way to use the cepstral swift engine in the same way, for me it would for fill all my needs.
Here is a forum post on how to make festival use better voices, but they were still not as good as the cepstral ones: http://ubuntuforums.org/showthread.php?t=751169
Then there are screen readers like orca, which maybe grate for their intended use to read the whole screen for blind people and must stay running in the back ground all the time. However orca to me is not easy to use and I have not found a way to set it up to read only what I highlight or copy to the clipboard. But it has built in drivers for both cepstral swift, festival and espeak.
KDE 3 seems to have had the perfect solution, with its ktts that could be used as a front end for festival, cepstral, espeak, and others. But in KDE 4 its been replaced with Jovie which in its current state of development can only connect to espeak. I also cannot find any information if its possible to run kde 3 apps on KDE 4 and also no where to download the old ktts. Here is a grate tutorial on how to set up ktts but as ktts is now obsolete its useless: http://awesomelinux.blogspot.jp/2009...-in-linux.html
Gespeaker can be used as a front end to some speech engines (espeak and other voices) but dose not proved good enough quality results. Also with some of the talkers it’s not possible to stop them once they have started reading.
This is a company (Digital Future ® http://www.digitalfuturesoft.com) who have a linux front end for cepstral which they say they can sell for $99.
“Dear customer
The version is experimental and it needs to be built and recompiled for ur version of Linux.
We can provide it for $99. Let us know if you agree and we will send you the secure payment link.
The adaptation process may take up to 2 weeks.
Let us know.
SupportDigital Future ® Team
http://www.digitalfuturesoft.com”.
I’m not shore about spending $99 on a program which is experimental and only dose what comes built into windows, mac and iso.
http://www.wizzardsoftware.com lso have high quality voices for Linux but out of my budget.
SVOX pico2wave https://launchpad.net/ubuntu/precise/+source/svox/ have a very light wait system which is quite customisable and can create many different voices. But not really suitable for reading back lots of text at a time it seems to crash and was not clear enough for my purposes. Here is the recommended bash script:
Code:
#!/bin/bash
pico2wave -l=de-DE -w=/tmp/test.wav "$1"
aplay /tmp/test.wav
rm /tmp/test.wav
The google chrome extension speakit uses tts from an online server to convert text to speech however this is also not a good solution for an expat living in China working over very slow internet connections: https://chrome.google.com/webstore/d...hrome-ntp-icon.
ispeech is similar and to me also unusable solution: http://www.ispeech.org/free.text.to.speech.tts.software.
JSpeak is a newer tts with a nice simple gui, but still lacks the quality I’m looking for.
So thats most of the relevant research I have done so far. I’m shore I have mist stuff out!… Please post any other solutions you have found below; all idea's are welcome.
I’m most interested in a bash scrip to pipe text into cepstral swift or if anyone has brought the Digital Future software before, was it worth it?
Thanks in advance, Eben
Bookmarks