It seems the Neo developers have made a new firmware chip capable of reading the track selection and navigation through the folders out loud.
Read more about it here

Now, I wouldn't wanna be caught dead with a Neo anymore, but I think this is a cool upgrade.

I know it's not voice recognition, more like text-to-speech, but it reminded me of the voice recognition option of the Empeg. I still hope it will become a reality one day!

If anyone gets one of these I'd be interested to know which TTS Engine they use.

We did this with a couple of open source engines but the quality is awlful. We then did it with a commercial engine which sounds excellent, but the small footprint embedded version isn't available until later in the year.

Rob

At the Amersfoort meet, didn't Patrick (IIRC) mention a text to speech chip that could be used (not 100% clear on how the sound output would be connected - separate output or mix with one of the normal outputs) on a third PCB in the tuner module kit?

/Michael

Man, would that be the coolest ever? It's been so long since we first talked about speech recognition. Not that I'm complaining or anything. I don't even regularly use half the features of the empeg anyway! Still, it would be a cool feature.

Text-to-speech is the conversion of written text into understandable english (or whatever language) sentences - i.e. making the computer speak.

This is different to the implementation of a voice recognition system. TTS speaks to you, voice recognition understands you.

It would still be a very nice feature.

HTH,

A.

Oops. missread the post

I was hoping that someone would take my challenge to make TTS on the empeg a reality, but all I really got was bellyaching that the quality of the speech wasn't good enough. I did a lot of the initial investigation and basically proved out the concept (using three freely available programs), but I really can't handle the integration of the TTS, sample rate conversion, and audio output into one binary without some help. All the ingredients are there, they just need to be combined and tweaked. We already have song names, artists, etc. in /proc/empeg_notify so I don't think it's a huge leap to suggest that if we got a TTS binary working, we'd easily meet or exceed what the Neo does.

Yes, but anyone who claims it's not going to be good enough is almost certainly right! Right now you have to spend money to get something worthwhile.

Rob

I guess our definitions of "worthwhile" differ. I think "worthwhile" is where I don't have to look at my display to get feedback from the unit. I don't need it to sound like someone sitting in my passenger seat.

In my testing I found that most of the things I threw at it were quite understandable, and the few that weren't could be remedied by having a phonetic equivalent stored in another field.

Anyway the thread I linked it looked like there was some thought of "demoing" the Empeg to Alan Black at CMU who writes Flite and its commercial-ware cousin Cepstral. If the guys living in Pittsburgh are still willing to consider that, I'd be happy to help out with any technical details or even help to set up a meeting, etc. I think we'd want to get a ballpark cost figure first though to see if it's even reasonable. I'll see if I can write him this week to discuss such matters.

Right now you have to spend money to get something worthwhile.

Yes, but what kind of money are we talking about? I certainly wouldn't mind paying a price for this. Ditto if the Voice Recognition option were ever to become a reality.

A Neo user posted a review of his experiences with this new Neo firmware speech chip. Here's what he said :

All right, fitting the new firmware chip is real easy, just pop the top of your Neo, prise out the old firmware chip and snap in the new one, takes just a few minuets. No real differences in the actual chip, same device type and memory size so I don’t know why they felt it necessary to ship it as a replacement chip rather than a flash update?

Powering up the Neo after fitting, reveals that the new firmware is rev 1.46, and the boot screen tells you that it is the speech version.

The Neo voice system is not really a voice, you have to install the supplied software on your PC, and after selecting a drive (your Neo), the software scans the directory structure and builds a small sound file for each folder and file on the drive. This file is saved in the same directory. So if you have 4000 tracks in 200 different folders you end up with 4200 small speech files on your neo disk, each file is about 11k in size. It’s these sound files that the Neo plays as you navigate folders and files.

I loaded the speech files into Cool Edit pro and played them with no problem, so I guess you could replace them with anything you want as long as the sound format and filename / extension are the same.

In use on the Neo it’s ok, although you do get some glitches if you navigate too fast. In terms of other features and bug fixes in the new 1.46 firmware it’s hard to tell, I listened to one of my play lists on the way to work with the random function turned on and it did play tracks which I’ve never heard the neo play before, weather this is just because the new firmware has a different random ‘seed’ or they really have improved the random function I wont know for a few more days.

The speech files are generated by some external DLL’s which the Neo guys have licensed from Fonix’s Ispeak, technology ( http://www.fonix.com/products/ispeak/ ) and do reasonable job although the female voice is too bassy for me, you adjust the speed and pitch of the voice but it really doesn’t make that much difference.

In conclusion, if you use the navigation controls on the Neo while driving then this is a useful update, however if you (like me) tend to select a play list or a folder before setting off, then Its usefulness is questionable however at 39 bucks with free shipping it’s a good offer.


So it actually generates and uses small speech files. Ugh. That's nice. Not. On my Empeg that would make about 8000 of those 11k sound files. Can anybody say : harddisk polution?

We need to implement a solution that generates the sound on the fly.
Where was the microphone for acceting input going to be on the Mark2/2a units?

The microphone input is right there on one of the wires coming out of the car sled. I'm still hoping someday it 'll be of some use.

this belongs in the tech section, but i just got my player back from repair and installed a compiler to start some empeg dev. i think there was some discussion on this board before about accessing the microphone via software?