View previous topic :: View next topic |
Author |
Message |
drolleman
Joined: 03 Feb 2011 Posts: 116
|
text to speech |
Posted: Wed Feb 29, 2012 7:35 pm |
|
|
I need to incorporate text to speech in my application. What is the best / cost effective chip / software out there ? Using google doesn't leave me with a warm feeling as specifications don't tell me what it will sound like. It seems phonetics are very important.
Does ccs use the dsp part of the 33f chips yet?
Is there a library? Third party? |
|
![](templates/subSilver/images/spacer.gif) |
temtronic
Joined: 01 Jul 2010 Posts: 9171 Location: Greensville,Ontario
|
|
Posted: Thu Mar 01, 2012 6:41 am |
|
|
best? lots of factors to consider,first is money! second is time!!
cost effective chip? don't use an external chip or module, cut your own code in the PIC.
software? Google will find 1,000s of hits,your decision what you choose
generally speaking ,not knowing how much you can afford or when the project is due,it's easier to buy a ready made solution,'off the shelf' that the PIC can access.
20 years ago I cobbled an 8bit ISA sound card to a 16F877 and it gave really nice performance, these days though I'd just buy a $5 chip.
as for the rest of the Qs.....don't know,for what part,yes. |
|
![](templates/subSilver/images/spacer.gif) |
John P
Joined: 17 Sep 2003 Posts: 331
|
|
Posted: Thu Mar 01, 2012 7:22 pm |
|
|
Do you truly need to read out arbitrary text, or are you using a set of phrases? If the data consists of numbers, you can encode them as short phrases which can be joined as needed, and there are chips and subsystems that allow you to record and play back the phrases. But decent text to speech algorithms that accept any input take large computers to run.
If you're willing to accept "robot speech" there is this:
http://www.sparkfun.com/products/9578
That chip needs an input corresponding to the phonemes that make up speech. If you want something to go from text to phonemes, you need an additional part:
http://www.sparkfun.com/products/9811 |
|
![](templates/subSilver/images/spacer.gif) |
drolleman
Joined: 03 Feb 2011 Posts: 116
|
|
Posted: Thu Mar 01, 2012 7:35 pm |
|
|
it will need to pronounce names. i will not know what it will need to pronounce at run time. some do this better than other chips. not many have a demo on their web site. so i was hoping that other developers could give me a idea as to what they have used in their projects. |
|
![](templates/subSilver/images/spacer.gif) |
temtronic
Joined: 01 Jul 2010 Posts: 9171 Location: Greensville,Ontario
|
|
Posted: Thu Mar 01, 2012 7:48 pm |
|
|
gee..hard to believe the '256' is still around ! good deal for $25 though you haven't said what you're willing to spend,number required,time you have for the project,etc.
I used the 1st generation 256 decades ago...with some tricks it sounds very very nice. |
|
![](templates/subSilver/images/spacer.gif) |
gpsmikey
Joined: 16 Nov 2010 Posts: 588 Location: Kirkland, WA
|
|
Posted: Fri Mar 02, 2012 12:31 am |
|
|
couple of things to consider here:
1) can we assume the language is English or are we talking (so to speak) multiple or other language ?
2) there is generally a fair amount of processing power required to generate reasonable sounding speech (other than pre-digitized stuff).
3) If you are dealing with names, that is a particular problem since so many people either pronounce their name different from someone else that spells it the same way or worse pronounces it in a manner that does not follow any expected rules (try figuring how Brett Favre is pronounced Farv !! )
My guess is it will take a fairly powerful PIC to be able to deal with some of those issues (and perhaps some learning (eeprom) ability to deal with funny names if you want to be correct)
mikey _________________ mikey
-- you can't have too many gadgets or too much disk space !
old engineering saying: 1+1 = 3 for sufficiently large values of 1 or small values of 3 |
|
![](templates/subSilver/images/spacer.gif) |
John P
Joined: 17 Sep 2003 Posts: 331
|
|
Posted: Fri Mar 02, 2012 9:20 am |
|
|
Maybe the operator of this site would be willing to let you use the database:
http://inogolo.com/
You could have an SD card filled with WAV files for lots of different names, to be retrieved and played as needed. I'm dubious about anything that tries to pronounce names from text. |
|
![](templates/subSilver/images/spacer.gif) |
drolleman
Joined: 03 Feb 2011 Posts: 116
|
|
Posted: Fri Mar 02, 2012 2:42 pm |
|
|
As long as the pronunciation is close that would be ok.
There can't be pre-recorded files as i won't know at run time what it will need to speak.
Yes, English will be the language.
Cost is fairly important as the device will need to be < $60. |
|
![](templates/subSilver/images/spacer.gif) |
SherpaDoug
Joined: 07 Sep 2003 Posts: 1640 Location: Cape Cod Mass USA
|
|
Posted: Fri Mar 02, 2012 6:53 pm |
|
|
Is there any chance the user could speak their name into it, and then just play back the recording? That would take care of pronunciation issues. _________________ The search for better is endless. Instead simply find very good and get the job done. |
|
![](templates/subSilver/images/spacer.gif) |
drolleman
Joined: 03 Feb 2011 Posts: 116
|
|
Posted: Sat Mar 03, 2012 2:26 pm |
|
|
The data comes from the database of thousands of names. Only a few are to be used per unit per day. The names will be uploaded to the device. so to have the a speech file would make the memory on the device huge, and time consuming to create. Most names will never be used but they have to be there. |
|
![](templates/subSilver/images/spacer.gif) |
|