1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
|
Here are the installation notes for the full system. Note that this
current version does not include any diphones. You should use the MBROLA
english diphones. They are available from http://tcts.fpms.ac.be
8/4/95 FreeSpeech Installation
Installation has been made as simple as possible, but may nevertheless
present some problems.
First of all, the system has only been tested on three different Unix
type systems. So for anything else, you're on your own. Having said
that the system takes a fairly simplistic approach to interfacing,
so as long as you have the basic ingredients, installation should
not be too difficult.
How much space? You'll need a minimum of 1 megabyte storage for
diphones and program. How much space it needs to run is determined
by the longest sentence you synthesize. With the default parameters
sentences of up to 100 words should be possible, (but a bit slow
:-) If you get the dictionary too, bear in mind the raw form is
about 8 megabytes, and the database form that is used by the program
is about 4 megabytes.
What is needed:
* An ANSI-C compiler
* A pronunciation dictionary. We tested with one called the Computer
Usable Version of the Oxford Advanced Learner's Dictionary
(CUVOALD). Because of the copyright restrictions it is not
contained in this package. It is available for ftp from
ftp.ox.ac.uk, in the directory /pub/ota/public/dicts/710 (or
similar). The same directory gives details of the copyright
restrictions. The program will still work without the dictionary,
but not as well. See the doc sub-directory for more information
about dictionaries.
If you ftp the dictionary put it in the sub-directory data with the
name text710.dat (the same as it is called in the Oxford archive).
And specify the name of your compiler in the makefile.
Then type make.
NOTE that you will get lots of compiler errors for certain files. This is
not serious. I didn't write them and I didn't want to correct them.
Note that building the dictionary database will take a while.
If you don't have the dictionary, or your system is not capable of
building a dbm database you will get a warning.
If the make process goes OK, you can type freespeech -o <output> to start
the program, and then enter text to be synthesized.
For use with MBROLA, use the -m option to produce the required format.
Or you could type:
freespeech -'?'
./freespeech -m -h ../dictionaries/lexicon| \
/home/stepheni/mbrola/mbrola /home/stepheni/mbrola/en1/en1 - - | \
sox -t raw -sw -r 16000 - -t raw -Ub -r 8000 - > /dev/audio
|
