Activity Team/gst-plugins-espeak
From Sugar Labs
|
gst-plugins-espeak
eSpeak library as a sound source for GStreamer.
Plugin uses given text to produce audio output.
Interface
gst-plugins-espeak is a valid gstreamer plugin thus it is a GObject
Properties
GObject properties:
- text text to pronounce
- pitch pitch adjustment, -100 to 100, default is 0
- rate speed in words per minute, -100 to 100, default is 0
- voice use voice file of this name from espeak-data/voices
- gap Word gap. Pause between words, units of 10mS at the default speed, default is 0
- trac track events
- 0 do not track any events (default)
- 1 track word events (see #Track words example)
- 2 track <mark name="<mark-name>"/> marks in text (see #Track marks example)
- voices read-only list of supported voices/languages
- caps read-only caps describing the format of the data
Events
Gstreamer uses separate threads and user should use gst.Bus messages(are processed in main gtk thread) instead of native GObject events. To use messages you need to setup tract property. These are supported gst.Bus messages (see):
- espeak-word before speeching a word, message properties:
- offset offset in chars from beginning of text
- len size of word in chars
- espeak-mark mark in text, message properties:
- offset offset in chars from beginning of text
- mark name of mark
Usage
gst-plugins-espeak generates raw audio/x-raw-int data.
Pipeline format
Plugin adds new URI scheme
gst-launch espeak://Hi ! autoaudiosink
Full pipline example:
gst-launch espeak text="Hello world" pitch=-50 rate=-50 voice=default ! autoaudiosink
Python examples
To use gst-plugins-espeak in Python:
- setup regular gstreamer envireonment
- plugin's name is espeak
- all writable properties(including text) make sense only at start playing; to apply new values you need to stop pipe.set_state(gst.STATE_NULL) pipe and start it again with new properties pipe.set_state(gst.STATE_PLAYING).
Simple example
import gtk import gst
def gstmessage_cb(bus, message, pipe): if message.type in (gst.MESSAGE_EOS, gst.MESSAGE_ERROR): pipe.set_state(gst.STATE_NULL)
pipeline = 'espeak text="Hello, World!" ! autoaudiosink' pipe = gst.parse_launch(pipeline)
bus = pipe.get_bus() bus.add_signal_watch() bus.connect('message', gstmessage_cb, pipe)
pipe.set_state(gst.STATE_PLAYING)
gtk.main()
Choir example
import gtk import gst import random from gettext import gettext as _
def gstmessage_cb(bus, message, pipe): if message.type in (gst.MESSAGE_EOS, gst.MESSAGE_ERROR): pipe.set_state(gst.STATE_NULL)
def make_pipe(): pipeline = 'espeak name=src ! autoaudiosink' pipe = gst.parse_launch(pipeline)
src = pipe.get_by_name('src') src.props.text = _('Hello, World!') src.props.pitch = random.randint(-100, 100) src.props.rate = random.randint(-100, 100)
voices = src.props.voices voice = random.choice(voices) src.props.voice = voice[0]
bus = pipe.get_bus() bus.add_signal_watch() bus.connect('message', gstmessage_cb, pipe)
pipe.set_state(gst.STATE_PLAYING)
for i in range(10): make_pipe()
gtk.main()
Track words example
import gtk import gst
text = file(__file__, 'r').read()
def gstmessage_cb(bus, message, pipe): if message.type in (gst.MESSAGE_EOS, gst.MESSAGE_ERROR): pipe.set_state(gst.STATE_NULL) elif message.type == gst.MESSAGE_ELEMENT and \ message.structure.get_name() == 'espeak-word': offset = message.structure['offset'] len = message.structure['len'] print text[offset:offset+len]
pipe = gst.Pipeline('pipeline')
src = gst.element_factory_make('espeak', 'src') src.props.text = text src.props.track = 1 pipe.add(src)
sink = gst.element_factory_make('autoaudiosink', 'sink') pipe.add(sink) src.link(sink)
bus = pipe.get_bus() jbus.add_signal_watch() bus.connect('message', gstmessage_cb, pipe)
pipe.set_state(gst.STATE_PLAYING)
gtk.main()
Track marks example
import gtk import gst
text = '<mark name="mark to Hello"/>Hello, <mark name="mark for World"/>World!'
def gstmessage_cb(bus, message, pipe): if message.type in (gst.MESSAGE_EOS, gst.MESSAGE_ERROR): pipe.set_state(gst.STATE_NULL) elif message.type == gst.MESSAGE_ELEMENT and \ message.structure.get_name() == 'espeak-mark': offset = message.structure['offset'] mark = message.structure['mark'] print '%d:%s' % (offset, mark)
pipe = gst.Pipeline('pipeline')
src = gst.element_factory_make('espeak', 'src') src.props.text = text src.props.track = 2 src.props.gap = 100 pipe.add(src)
sink = gst.element_factory_make('autoaudiosink', 'sink') pipe.add(sink) src.link(sink)
bus = pipe.get_bus() bus.add_signal_watch() bus.connect('message', gstmessage_cb, pipe)
pipe.set_state(gst.STATE_PLAYING)
gtk.main()
Simple TTS example
import gtk
import gst
import pango
window = gtk.Window()
window.connect('destroy',
lambda sender: gtk.main_quit())
workspace = gtk.VBox()
window.add(workspace)
# text widget
scrolled = gtk.ScrolledWindow()
workspace.pack_start(scrolled)
text = gtk.TextView()
scrolled.add(text)
buffer = text.props.buffer
buffer.props.text = file(__file__).read()
tag = buffer.create_tag()
tag.props.weight = pango.WEIGHT_BOLD
# play controls
toolbar = gtk.HBox()
workspace.pack_end(toolbar, False)
play = gtk.Button('Play/Resume')
play.connect('clicked',
lambda sender: pipe.set_state(gst.STATE_PLAYING))
toolbar.add(play)
pause = gtk.Button('Pause')
pause.connect('clicked',
lambda sender: pipe.set_state(gst.STATE_PAUSED))
toolbar.add(pause)
stop = gtk.Button('Stop')
stop.connect('clicked',
lambda sender: pipe.set_state(gst.STATE_NULL))
toolbar.add(stop)
# gst code
pipe = gst.parse_launch('espeak name=src ! autoaudiosink')
src = pipe.get_by_name('src')
src.props.text = buffer.props.text
src.props.track = 1 # track for words
def tts_cb(bus, message):
if message.structure.get_name() != 'espeak-word':
return
offset = message.structure['offset']
len = message.structure['len']
buffer.remove_tag(tag, buffer.get_start_iter(), buffer.get_end_iter())
start = buffer.get_iter_at_offset(offset)
end = buffer.get_iter_at_offset(offset + len)
buffer.apply_tag(tag, start, end)
bus = pipe.get_bus()
bus.add_signal_watch()
bus.connect('message::element', tts_cb)
# gtk start
window.show_all()
gtk.main()
Known issues
- espeak-word with espeak < 1.40.09 doesn't track words with numbers(at least full-numered words) in proper way
- if you are tracking espeak-word with espeak < 1.40.10 you should use gst-plugins-espeak-0.3x(or 0.3 branch in git repository)
Install
gstreamer-plugins-espeak is a part of SugarPlatform-0.84, so just install meta package(depends on your distro) with SP.
XO
- attach [1] repo(or just download proper .rpm)
- install gstreamer-plugins-espeak package
Contacts
- Aleksey Lim
- be involved and add yourself here
