> Some comments from me then:
> * You have used "A" as a command. That is not possible and should not work for
> you. The letter A is reserved for the "DTMF anti flutter mode" as described in
> the SvxLink Server user documentation:
Ok, shall I use commands beginning with "0" as help, f.e. "01#", "02#", ...
I think it would be well to provide different help sections (f.e. "01#"
plays the handling of the module, "02#" plays the predefined airports,
and so on)
> * Try to reuse already existing sound clips from the "Default" directory.
> Especially numbers you should not have to use your own. There already is a
> function "playNumber" that do all the dirty work of playing back numbers as
> intelligent as possible. The playNumber function have been recently improved
> so you should be able to use that.
Ooops, I think you can't use existing sounds in terms of C++ subroutines
if it made by different people. If you and I speak the same its always
very specific and different. I've tried out to take the numbers you sad
and add my personal sounds....the result of course is a switching
between the voices that sounds very strange. Yes, it's working from the
view of a programmer ;-) but in my opinion the sounds shouldn't change
So I think the way you want to go (mary) is effective for further
programming, but I will miss the personal touch of your voice.
> * When using processEvent, try to think in larger events and not in small
> chunks as words, if at all possible. It feels a bit wrong to have an event
> function called "say" and then call "say moderate" or "say clouds". This can
> make it harder to make a translation if the words have fixed positions in the
> announcements. As a comparison 24 is "fier und zwanzig" in German and "twenty
> four" in English....
> I don't know if it's possible to change the code and I don't know if it's a
> problem. Honestly, I have not studied the code thorough enough to understand
> it. I just felt I should comment on it :-)
I'm not happy with the code, yes it works, but it's not optimal. I'm
searching for a more effective way, a metar-parser for example that is
There is a metar C library I've used as a template, they have gone the
same way with a lot of "if...then"'s.
The main disadvantages of the Metar-strings are the high numbers of
exceptions, especially in the RMK section (after QNH). The RMK section
is country specific and the weather men put in what they want ;-(
> Speaking of sounds, I have changed my approach a little or rather been able to
> do what I really wanted to do from the beginning. The sound clips are now
> generated by a text-to-speech system rather than having me record them. I have
> been able to do this since I found a text-to-speech system with good enough
> quality. The one I found was MARY (http://mary.dfki.de/), a German
> application. I'm sure you're glad to hear that ;-)
It's not important who has written the application, for me it's only
important that it's open and free and you can use it without facing the
charge by a lawyer. I've already used mbrola (that is a part of them) a
couple of years ago and have generated sounds for svx with that too. It
was very interesting to play with the different voices and languages ;-)
> There are many benefits of doing this:
> * No Swedish accent
I think the people like them and will miss it.
> The drawback at the moment is that it can be hard to hear what is being said
> in some situations but I'm sure that can be improved upon.
> Now, Adi, it would be really interesting if you could spend some time on
> reading the README file under svxlink/scripts/sounds/mary and see if you can
> generate the sounds for yourself. Then you can add your own sound clips for
> the MetarInfo module and everything will sound the same!
> Isn't that nice :-)
Yes, it is ;-) I've already installed mary.
73's Adi, DL1HRC