I can get one voice to be 100% accurate on controls (as their blog states), but the next time it starts up in the same conditions, it may be 40%. And when moving, I speed up, 20%, slow down, 80%. They are not ready for retail yet. I also haven't heard anything from them in several months.
The built in voice control for StreetDeck is much more stable as it uses microsoft teaching for capture. One Voice doesn't. I showed it off at a meet in Dallas (one voice) and on my voice it was ok, but the other person, it wasn't nearly as accurate. Don't get me wrong, it's a really nice toy and shows great promise, but not really 100% functional. It needs the teaching even if One Voice doesn't want to do it.
Hey thanks for creating that video. It motivated me to get voice control for my carpc in a hurry. I always thought it would burden my CPU if running constantly, but seeing the "Speak prompt" I realized this isn't the case. This is excellent because I can use a steering wheel button to trigger voice activation and minimize any potential CPU burdens.
I looked into the Andrea MICs and the automotive version looks really nice but the price is too high ($320!), so I settled for and ordered the "
USB-SA Andrea External USB Soundcard / SuperBeam Microphone Bundle " which can be found here: http://store.mp3car.com/External_USB..._p/aud-004.htm
To get myself familiarized I grabbed an old microphone, installed Road Runner on a test computer, downloaded SAPI 5.1 and got it working on my first try. The Road Runner version works but is not as fancy as the Street Deck version. I was almost ready to switch over to Street Deck after seeing your presentation. LOL.
Which text to speech voice were you using in that video?
I have voice working quite well with Road Runner. What I can NOT get it to do is manage my music. EA: "Play Artist The Doors" ect
I just do not know how to create the grammer file for something like this. Anyone know how to do this? With thr RRMedia database, it is capable of exporting a text file with every song, and the path to it. If I know how to create the grammer file, I could easily write a small app to create the grammer file in the correct syntax from that text file.
The key is to have some sort of keys in the grammer file for:
[Play artist xxxx]
[Play Song xxxx]
The problem I am having is to get MS Voice recognition to use just the phrases that are defined in the grammer file, and not every word from every song ect.
For instance, if you have a song songs like:
Tomorrow today and yesterday
tommorrow is the day
yesterday is so far away
you could speak "tomorrow is far away" and it would recognise it as a phrase, but not play anything, as the words are all in different titles.
If anyone out there is familiar with creating a grammer file, I would love to hear back, and see if we can come up with something!!!!!
You have me so confused now! LOL! You mention above that One Voice does not work that great. I PM'd you about it, and you basically said the same thing, that it does not work that great. The program you link to above is also from One Voice, and I would imagine uses the same VR engine / coding.
That prog also is made for itunes and WMP. That would prevent any interaction with Road Runner and Winamp.
That program is pretty cool However it does not seem to practical for RR control.
The bottom line, is just being able to say "Play Artist The Doors" and have RR then play all music by The Doors.
I have researched this for literally months now, and have not come up with a solution. Anything that is out there is designed to work with iTunes or with WMP, as it is based on their libraries of music.
This is something that I think is going to have to be put together from scratch for the most part to work with Road Runner. Blue's RMMediaDB can export the list of music in your collection, and I can easily parse that to get artists, songs ect. From there it is creating a Voice Recognition app that can listen to something in your library being requested, and then carry out the command to send to Road Runner.
Wow! That would be great! I have tried in the past over and over and over to come up with a solution to mimic the Ford Sync capabilities.
As for the RRMedia path dump, check with BlueZX3. I just ran the one I had here, and it just exports each album, and not the full path to the songs ect. I know he has a copy of one that does, I just can't find it here
Have I got a suprise for RR users! I have a ton coded up now, running 24/7, no memory issues.
Here is what it does so far.
Can do any command that can be exec'ed.
Uses ini for all commands matched to flat text file.
Each skin can use its own text file for commands or use master file.
Each text file is loaded dynamically into memory. This narrows command match and is extremely fast.
So, when you say, Computer Select Audio, it goes to your Audio Skin and loads a skin specific command set. If no command set is configured it stays with the master one.
Like when you say Set Volume Twenty, no command set change is needed, just the command SETVOL;Master;20
Configurable preCommand. Computer or whatever.
Standalone exe using sendmessage. Hides in tray
Easy ini and text file command loading.
Just load RR and SkinTool and run through menus picking commands and choosing own phrases to match on. Paste or type in, Click add and go.
I'm using SAPI 5.1 and the recognition is very good.
Using small command sets really speeds it up and improves accuracy.
Might just size it for embedding. Just have to code the Twips to Pixel bit.
I am now on the mic mute control, although I have had the volume way up and it still matches commands.
Shoot me some feedback on other stuff to add, ideas, ect...
I'm stoked, cause like everyone else, I want this, and there has been no follow through on Voice Control for RR and I think it is needed.
I looked at all of the existing and nothing works correctly.
I have approached this with the idea of fully configurable command sets, user choices, and ease of use.
You are not going to "set up RR" by voice control while driving.
You only want the usable functions of RR and your skin of choice to voice command. Keeping it tight with only the commands you want in each skin, will let you remember the commands to say easier.
Here's a pic of it what I have right now.