Hi Felipe.On Mar 4, 2011, at 6:02 AM, Felipe Castro wrote: Hi David,
2011/2/27 David Hill <address@hidden>
Thanks for the two emails regarding gnuspeech, including the information about installing on a Debian system under GNUStep, and the useful links. Interesting that you had an undefined variable "false" in the "Synthesizer" compilation. Given that "Synthesizer" is basically a very initial version, I suppose I should be pleased there were not worse problems. The App still requires a fair amount of work -- graphs that are simply stubs and that sort of thing, uncontrolled creation of unused file handles until it crashes in certain conditions, and a lot of removal of redundant, unused code . But it does run correctly enough to give a feel for the tube model, and how the output varies as you vary the basic parameters. These things are on my "to do" list.
Yes, ok, I could feel what the Sinthesizer is, I just didn't feel comfortable with the spectral analysis, because I felt the resolution not so good. Well, I know I should practice it much more... Maybe it could be thought of visualizing the generated waveform too.
The resolution is decided by the bin size you choose for the analysis, which you can vary to suit your needs. As for the waveform, there are other tools that can display the waveform, but it could be a useful feature to include in the App. I'll give it some thought. My initial reaction is that it wouldn't add anything useful to the function of "Synthesizer" as the output at any given time is constant. But, given that the waveform is repetitive and available, it would be fairly easy to add such a display for just a few cycles of the repeating waveform. What would you want to see by looking at the waveform? Comparison with real speech? ???
So, are you working in your local repository without updating the remote one (at Savannah)? Because I see the last update there dates from 2009 (am I wrong?).
You are wrong. There is no material in the "Download" area of savannah because there is no official release, but the most up-to-date material is in the repository. It just happens that we have been working on other things for more than a year. The dates are correct, and represent the latest work. An official release is coming shortly, and will simply copy the relevant parts into the download area.
When someone works on a particular bit of code, they do indeed check out that code, work on it, test it, and then check it in again using SVN (the CVS repository is no longer in use) but that is normal for a group project.
Work should continue, and the Download area should be updated each time significant additions have been incorporated and tested.
Please keep us in touch on how you make out with the various components. There's an official release of the software coming up soon so your input will be very useful.
Ok, I have read the docs about the process and understand the general approach. I'm very interested in the articulatory method, because we don't see it in other part, at least in the free software world. I'm brazilian (sorry my poor English), and here people seems to be using only contatenation and formant synthesis. And my first goal is not related to Portuguese, but Esperanto.
Any good documentation concerning the compilation and use of the suite of software (Monet, Synthesizer, TTSServer, etc) under Debian GNU/Linux GNUStep would be of great interest.
I have found some annoying bugs in the GNUStep system here, like unresponsiviness of some widgets, but maybe it is related to the alpha state of gnuspeech.
Anyway, I'll try to document my efforts, hopefully it will be useful for someone else.
In order to study ObjC and GNUStep, I tried to launch the Project Center, but it was stuck at an unusable way. Today I'll see it again, in order to test if the infamous machine-restart works for me. :-)
I also installed WindowMaker, but no success then.
Your comments on difficulties with GNUStep are fair, and it is an ongoing problem that has slowed development on the GNU/Linux version of gnuspeech.
You are probably aware that the whole system is still under development. Monet needs the editing stubs expanding so that it can again be used to develop the databases needed for different languages (the tools, such as Monet and Synthesizer, in the software suite, are what we used for the current databases that provide TextToSpeech for English). Marcelo Matuda did most of the port from the Mac version to the GNU/Linux version,but I haven't heard from him for a while.
Yes, I tried to use Monet, but didn't get much of it here.
So you work on OSX firstly and after that update the GNUStep part?
So, if sometime I succeed to help on the code, I should first wait for the updates in the OSX, to port them to the GNUStep?
Or, to start with, you could simply bring the GNU/Linux-GNUStep version up to the current Mac OS X standard. That would be a good start. This is a topic about which I should very much like to hear from Marcelo Matuda, who did all the work on the current GNUStep version.
There is a fair amount of documentation to provide understanding of the system. Please let me know if you have any questions.
I didn't go through all the documentation of Monet, because I couldn't play with it. But I noticed differences in the interface for the new port. Just couldn't see any Inspector window, for example. Is this such a stub to be reimplemented?
Which "Inspector window"? There are twelve different tools/windows that are available in Monet. The ones concerned with the synthesised speech are complete. The ones concerned with editing/creating new database components are stubs. But none of them should be called an "Inspector window".
I wonder which documentation you have read.
When you say you couldn't play with Monet, do you mean it wouldn't run at all, or simply that the editing facilities were absent?
If you have access to a NeXT system you can install the NeXT version of the TextToSpeech kit on that and see how the complete system works. It is useful for assisting in the port, since any aspect can be checked directly with a complete, bug-free (?) system.
Unfortunately, it will be very difficult for me to see that kind of hardware, here in Brazil... :-)
Such hardware is available at low cost from BlackHole Inc. in the US ( http://www.blackholeinc.com/ ). No doubt it could be shipped to you, but I can see the problems you might encounter.
Well, I'll try hard to understand that code, because I really wants to play with the system... But for that, I have first to take a look at ObjectiveC tutorials, because what I only know for now is C.
If you have access to a Macintosh running OS X, you could get a good feel for Objective-C by working through the examples in Aaron Hillegass's book, which I strongly recommend:
Cocoa Programming for Mac OS X (third edition) (Addison Wesley 2008, ISBN 978-0-321-50361-9)
GNUStep is intended to be very compatible with xcode/Interface Builder on the Mac (they have a common origin as you know).
Someone else on this list may have better comments, and a better recommendation for getting up to speed in Objective-C and GNUStep.
Hope this helps.
Warm regards.
david ---------- David Hill -------- The only function of economic forecasting is to make astrology look respectable. (J.K. Galbraith) --------
|