Thread: Text to Speech Chip/Module

I watched this vid a while ago, not sure if it is the same thing you are looking at

The GPS does not use a TTS chip, it uses a TTS engine, usually Loquendo, Nuance or the older Vocalizer voices.
Similar story on mobile phones, so maybe it would be easier to find/progam a software for Android that accepts a serial input or file transfer and speaks out the result using the installed voice engine.

I'm trying to avoid that robot voice B4L.
While the emic is ok for simple TTS aplications, it's really crap for the application I need TTS.

Thanks Downunder, I don't care if it is a chip, module or engine.
What I really need is a black box which will accept serial RS232 and spits out a sexy female voice. Pretty much that simple.
Though it needs to run on 12VDC and has to be stand alone. If I could make a zombie GPS and just utilise it's voice function, then I'll do it.

If I have to load phonomes onto something like a PIC chip, I'll do that can construct my own words from syllables etc. (obviously this isn't as easy as it sounds or I would have done it already)
Or if I can find software for something like the raspberry pi then I'll try that. (I'm currently running espeak which slightly better than the emic, but still a little robotic).

So re-purposing a GPS platform might work, if I knew what that platform was or even it's OS.

Emic 2

While the emic is ok for simple TTS aplications, it's really crap for the application I need TTS.

If it sounds like robotic, then it's not appropriate for the job I intend to use it for.

Lets just say I'm going to use it for a sperm extraction vending machine.
Maybe something like this...
So I need a pleasant sexy sounding female TTS.
Getting your gnob gobbled by Stephen Hawking just isn't going to work.

How about , or , like the Nuance (former LoquendoTTS) products all these can be used on embedded systems.
Of course a bit of work if needed to implement it into your program, but it might be easier to do it that way as you can use the same platform for the other tasks as well, not just the text output.

Works for me Trash LOL

If you do go down the GPS route; Garmin have free Voice Studio Software for creating personalised voices with a microphone.
Not sure if you can customise it enough for your needs though.
Also will require a compatible GPS to operate.

Thanks again downunder.
Cereproc supports raspberry pi, so there's a good chance I can use that provided they're not going to cost an arm and a leg.
The other two don't appear to be of any use. It's a major pain in the arse to have to pick up another platform unless that platform is a black TTS box. (12V in, V.24 in, Audio out)

I'm reluctant to work with the rasberry pi when all I wanted was an affordable (non robotic) TTS module.
But the pi is cheap enough to make it worth a little bit of effort. I struggle with linux and python (not sure why) but I can manage and maybe even move my PIC control functions onto the pi.

I'll see how much the Cereproc ends up costing, because ripping the voice box out of a GPS could end up being cheaper.

I heard a female robotic voice via a UHF repeater today. Even though it still had a noticeable robotic quantisation noise, it was quite acceptable to listen to even over a radio channel.

I tried this on my Samsung S3 but not to sexy LOL

I don't know how good your coding skills are but there is a way for Python to make it speak:
Depending on your skills it should also be possible to write a little Android app that accepts text over the USB or WiFi connection and and output using one of the avialable TTS engines.
Basically like an Ebook reader with text input on demand.
Did not do much searching on this but there is a chance projects for this already exists.

My coding skills are pretty ####ing awesome. (pat myself on the back) but I find because I have not exercised at lot of them I have lost of lot of skill in the higher level languages
and my 32bit program has never been any good. The bus stopped at 16 bits and I got off.
My main programming is with assembler and I can shit it in if the operating system doesn't get in the way.
I'm a bottom feeder when it comes to low level and risc programming.

So ideally, what I wanted to do was run a PIC chip which I can program in assembler and talk to a TTS piece of hardware. It >should< have been that simple.
But without any TTS with a good sexy voice I have to compromise. The raspberry pi was that compromise and the thought was just to use the pi as a TTS module since it's about 10 times cheaper.
But TTS on the pi is turning out to be a crock of shit. It doesn't exist and it's just as bad as a $5 speakjet.

So, back to a risc mentality. While TTS would have been ideal, I can live with much less. What I need the TTS to say can be broken down into about 50 blocks of wav files and
they can be strung together sequentially using a simple script like perl.

Anyhow.... I have had another look at python. In days gone by I would have picked up this language in about 30 minutes. I must be getting old because it's kind of confusing.
I struggle for the time, but I expect this will change. I've also had to go back and look at Perl.
So at the moment I'm flipping between python and perl trying to work out which I can program with the least amount of fuss.

What I'm currently running is MPG321 on the pi and it is playing the TTS generated and captured files as MP3 files. It works really well.
The problem is now I have to make it work in perl or python. Which is where I'm at.

Then I can port up some of the functions I would have had the PIC do. So in python it might check a GPIO for data input and then assemble the TTS phrase based upon that data.
The actually programming of how to assemble the speech is a piece of piss. It's getting the data from the port and making perl or python run MPG321 and pass variables to it.

I'm working on it now and I'm always open to tips and shortcuts if anybody would like to suggest some.

Many years ago, when the commonwealth games were in Brisbane, the ABC made a number of digitized speech devices for line identification.
If you only require a limited number of responses you could get a female of your choice to say them and record / digitize them in WAV or MP3 and then play them back as required.