About the TTS Engine
AT&T Natural Voices Text-to-Speech (TTS) Engines provide synthesis services in multiple languages for xMatters Voice engines on version 5.0.
Hard Drive Space
- The AT&T Natural Voices server requires a maximum of 50 MB of disk space (without voice fonts).
- The disk requirement for voice fonts varies depending on the type and number of fonts you install, but as a general guideline, assume between 150 and 200 MB of disk space for each 8 KHz voice font.
- You will also require some additional space to store the server logs.
- The server requires a minimum of 256 MB RAM (recommended 512MB). Extra memory will provide better performance, particularly if you are using the TTS engine while other applications are running.
- Each voice requires approximately 75 MB of data to be accessed by the server. It is recommended that you have at least 1 GB of memory per server CPU if you start more than one voice.
- The memory requirements increase based on the number of channels: each additional channel requires an additional 10 MB RAM.
You can store the AT&T Natural Voices in any directory, but this article assumes that AT&T Natural Voices reside under /usr/local.
The tar file will unpack into a single directory called att_naturalvoices_v5.1_server. You may safely unpack the tar file in any directory.
To extract the files from the tar file, use the following command:
tar -xvf att_naturalvoices_v5.1_server.tar
Install Voice Font(s)
Before attempting to use the server for the first time, install one or more voice fonts. Each voice font is packaged separately, and its tar file will unpack files into a single directory. To install a voice font, simply copy all the extracted files into the appropriate directory.
For example, to install Crystal 8 KHz (US English):
tar -xvf att_naturalvoices_v5.1_crystal8_pcm.tar
cp crystal8/* /usr/local/att_naturalvoices_v5.1_server/data/en_us/crystal8
To complete a voice font installation, you will need to adjust the /data/tts.cfg configuration file by uncommenting the font that you have installed. Commented lines have double pound (##) in the beginning of the line, while uncommented lines have a single pound (#) in the beginning of the line.
Continuing the above example, you would need to replace the following lines:
With the following:
Using the Server Edition
The AT&T Natural Voices Server interacts with xMatters using a client/server architecture where the xMatters application and Natural Voices server run on the same or different computers.
Before running the server, you will need to change the permission of the TTSServer so it can be executed; i.e.:
chmod 744 TTSServer
Running the Server
|Required field. Identifies the port on which TTSServer will be listening for client connections. Note that the client application must be configured to use this same port.|
|Path name||The directory path of the data subdirectory. Default path is $NaturalVoicesPath.|
|Voice font||Specify the voice: mike8, crystal16, etc.|
Trace level 0
|Prints only error messages.|
|-v1||Trace level 1||Prints error messages, voice information, and process start/stop information. This is a default.|
|-v2||Trace level 2||Prints error messages, voice information, process start/stop information, and request and reply packets. Beware that v2 mode generates extensive information and should be used sparingly.|
|-v3||Trace level 3||Prints error messages, voice information, process start/stop information, request and reply packets, and all notifications. Beware that v3 mode generates extensive information and should generally not be used.|
|-config||Configuration file||Default is /data/tts.cfg|
|-m||Number of clients||Maximum number of simultaneous clients. Attempts to allocate additional channels will fail. Default is 32.|
|-y||SNMP port||Access this port for SNMP.|
|-temp||Path name||Use the given path for temporary files (default is /tmp).|
|-if||Interface||Bind to specified network interface.|
|-wd||#||Allow # timeout intervals before disconnecting client.|
|-l||Log file path||File where messages are logged. Default is stdout.|
|-rollover||N [unit]||Rollover log file after N minutes (or specify unit of time (hours, minutes, days, weeks)).|
Specify the path to voice font at command line (-r):
/usr/local/att_naturalvoices_v5.1_server/bin/TTSServer –c 7000 -l tts.log –x mike8 –r
Specify the path to voice font in an environment variable (NaturalVoicesPath):
/usr/local/att_naturalvoices_v5.1_server/bin/TTSServer –c 7000
Generated Output Messages
Server Output and Error Messages
The TTS Server writes all output and error messages to the log file specified with the –l option. All messages have the following format:
where is a sequential number of speak requests (each speak request creates a new child process) and
Wire#1 171.90 : wireline started
The following messages are generated by the TTS Server during initialization. These messages indicate that a wireline server started up.
- Wireline is the application itself
- server engine is the thread that receives wireline requests and sends wireline replies
- the asynch notify server is the thread that sends wireline notifications
- the asynch standalone is the thread that processes speak requests.
Wire#1 171.90 : wireline started
Wire#1 171.91 : server engine started
Wire#1 172.81 : asynch notify server started
Wire#1 172.85 : asynch standalone started
The following messages are generated by the TTS Server during shutdown and map directly to the corresponding Server Initialization Messages.
Wire#1 177.56 : asynch standalone finished
Wire#1 177.56 : asynch notify server finished
Wire#1 177.56 : server engine finished
Wire#1 177.60 : wireline finished
Voice Inventory Messages
The following messages enumerate the voices that are available on the TTS Server, the default voice and the current voice:
Wire#1 172.93 : Installed voices
Wire#1 172.94 : Default voice: crystal8;en_us;female;adult;8000;16;pcm;
These messages may be displayed when xMatters disconnects from the TTS server.
Wire#1 368.83 : receive error 10054 ::Connection reset by peer
Wire#1 368.83 : sending notify packet::Connection reset by peer
Wire#1 368.90 : wireline finished::Connection reset by peer
Periodically, the server will clean up resources for child processes that have exited:
TTSWireServer 395.02 : reaped child 1
These messages are simply housekeeping messages and can be ignored.
Once you have installed the AT&T Natural Voices software, you can configure the xMatters (alarmpoint) engine to use the new TTS engine.
To configure xMatters to use an AT&T Natural Voices voice:
- Log in to xMatters as a Super Administrator, and navigate to your SIP Device Engine details.
- On the SIP Device Engine details page, in the TTS Engine drop-down list, select Natural Voices.
- In the TTS Engine Personality field, type Crystal
- In the TTS Engine Server field, type the IP address of the server running Natural Voices.
- In the TTS Engine Port field, type the port number used for the Natural Voices server, 7000 in the example above.
- Click Save.
Originally created by Craig Gulliver