IN PROGRESS: Fixing up Samantha

[Yettavr6]

If anyone is interested, i'm working on making Samantha a little more natural sounding and speaking proper US English. Hopefully no one has already done this. If so I appologize. If not, we should add this to the sticky once its actually done and uploaded.

 

So far what i've done and tested:

-Replaced "roundabout" with "traffic circle"

-replaced "turn back with "turn around" ...also considering changing it to "make a U turn"

 

Just now i've done the following but haven't tested. Will test on my way to work tomorrow and make changes as needed:

-slow the voice down to make it more natural. The demo clip of Samantha on her creator's website (Nuance) is slower and sounds much more natural

-change the way highway numbers are pronounced. e.g. instead of Route 422 being "four hundred twenty two" it will just be "four twenty two"

 

Once i test this and know that it is all working I will upload it as the entire zip file that can just be reimported using the directions on Pioneer's website for 2.0

[jneilon]

That is great. I would love to know when you have it ready for download

[peterparker]

if you could make her less robotic i'd actually use TTS

[skicrave]

This is one of the biggest complaints I have with the actual navigation component of the system. I don't even use TTS because it sounds so terrible.

 

Have you thought about adding another voice, that may be easier in the long run?

[peterparker]

i've tried looking for other voices... but haven't looked that hard yet..

[Yettavr6]

I found a voice called Susan on an iGo forum, that is supposed to be the clearest one available, but I haven't been able to get it to import yet. It looks like the filesystem structure on the F is different from a standard iGo device since the VR uses Samantha as well. The voice i downloaded has a folder called "tts_loq" that goes in the iGo directory, whereas the F has a folder called "ScanSoft" instead (Googling ScanSoft is where i found Nuance). Then the files inside these directories don't match up either. The voice has a couple of .lex, .vde, .sde, etc files, whereas the F has a ton of .dll and .dat files. At best I could probably get iGo to use the new voice, but VR would still use Samantha. I'll just need to figure out what file tells iGo which voice to use. After hearing the sample of Samantha on the Nuance website though, I'm convinced she could sound better with the speed turned down and a few other tweaks.

 

Also, something else i forgot: I'm also going to reduce the amount of prompts when using the TTS voice, but i need ideas of what the prompts should be. Right now for most situations there are prompts at 50, 100, 300, and 800 meters OR 50, 200, 500, 1000 meters. Some highway situations go out to 2000 meters. I'd like to reduce this to 3 prompts total i think, but i'm not sure which distances would be good. What do you guys/girls think?

[Yettavr6]

OK, i had a little progress today, but not really as much as i had hoped. Here's what i did:

 

-I mentioned in my original post that i had changed "roundabout" to "traffic circle" and "turn back" to "turn around". I've now changed "turn around" to "make a U turn". These have all been tested and work.

 

-I also mentioned that i slowed the voice speed down (from 50 to 40). This makes the voice slightly less irritating.

 

-I found a setting for each prompt type, that specifies a time period that must have past since the last prompt for that prompt to play. I changed some of these so that in many cases there will only be 2 prompts to turn instead of 4. If you are in the city stopping at alot of lights, or stuck in traffic, you will get more prompts then since the time period is more likely to pass. I think this is the best way to handle reducing the prompts. I tested this on my way home from work, and typically only received prompts at 1/2 mile, and make the next left/right, whereas on the way to work i would get 1/2 mile, 1/4 mile, 400 feet, and next left/right. The time period is also per maneuver so if you have two turns within a few seconds, you will still get the prompts for both.

 

-I found a 22khz version of Samantha from MioMap (ours is 11khz) and successfully got it working on the F after much trial and error, but it really doesn't sound that much better. It is slightly noticeable, but not the results I had hoped for. My VR also stopped working.

 

All of the changes I have made so far are in the commands.csv file within the TTS zip file. I will post this zip file in a few minutes for anyone that wants to try it

 

For the 22khz voice, you need to dump a bunch of files in the iGo\ScanSoft directory. I don't really want to post this yet until i figure out why the VR broke.

 

Also, there is a rules.ini file in the TTS zip file that defines how highway/exit numbers should be pronounced, but its not being used for some reason. Based on this file, 422 should be "four twenty two" and 202 should be "two oh two" but both are said with "hundred". I need to figure out why this file is not being used because it really annoys me being told to enter highway four hundred twenty two.

[skicrave]

Thanks for the update.

 

Here are some TTS voices I found: http://www.filestube.com/13f74da11125c5 ... tails.html

 

It may be worth trying to adapt those to work on the F-Series, but I haven't been able to figure out how to make it work.

[mikeinnj]

Keep me posted YettaVR6 on modding the voice. My only and I mean my ONLY complaint with the unit right now is how the voice never seems to STFU with the turns. It sounds like you did EXACTLY what I want to have done with the voice. Right now I have the mute button mod which is a beautiful thing, but it would be nice to not have to hear about my turn every 10 feet. Awesome work man!

[Yettavr6]

Attached is what i have so far. This will give you:

 

-Traffic Circle and U-Turn mods

-Reduced prompts

-Voice speed reduced slightly

 

To install, just put the zip file in a TTS folder on your SD card or USB memory stick, and follow these instructions from Pioneer:

 

10-1. Touch the button to the right of "Voice Language" and then choose "English (US) (Jason)" from the list.

 

 

 

10-2. Touch the button to the right of "Voice language" (which shows "English (US) (Jason)") again.

 

 

 

10-3. Touch the "TTS Setup" tab on the right bottom of the screen.

 

 

 

10-4. To update English TTS file, touch the "English (US) (Samantha):TTS" button.

 

 

 

10-5. Touch the "from SD" button.

 

 

 

If you cannot select "from SD", please check your system and SD card as follows:

 

Check your SD card to confirm that the SD card contains the correct folder and files in the correct format.

 

10-6. Touch the "English (US) (Samantha):TTS" tab. Then English TTS file will be updated immediately.

 

Try it out and let me know what you think about the frequency of the prompts. The voice quality is not really improved yet so don't expect much there

Voice_TTS-nua-eng-us-f1.zip

[BorisM]

While you're at it, there's another things that bugs me (haven't seen anyone else reporting it). If you set units to miles/yards, she will pronounce distances as "do something in 2.5 mile ... ess"

 

I've traced it to distances_yd.csv file, what she is supposed to say is written as "mile,s" in the feet version, of course, it is written as "miles".

 

Beats me how these things got through quality control...

[Yettavr6]

Good catch. I hadn't even tried to use yards yet. The attachment above now has that fixed.

[BorisM]

Thanks. I tried using yards once (yards are closer to meters ) but constant ...ess!!! drove me nuts.

[1fastrsx]

Tried using this tts and it was going REALLY Fast... back to Jason for now.

[Yettavr6]

DOH! I forgot to change the info.ini back to 11khz. I just fixed it and uploaded again