Generating speech stream from text via RHVoice library without re-initializing engine. This very fast and more convenient than call RHVoice-test.
Supported audio formats: wav
, mp3
, opus
, flac
and pcm
(raw RHVoice output).
pip3 install rhvoice-wrapper
This package does NOT provide RHVoice itself. You must be build (or install) RHVoice, languages and voices manually. In Windows you must specify the paths for work.
Warning! rhvoice-wrapper-bin
not working in macOS, install RHVoice manually.
Instead of RHVoice you may install rhvoice-wrapper-bin. This is best way for Windows.
If the rhvoice-wrapper-bin
is installed, its libraries and data will be used automatically.
pip3 install rhvoice-wrapper[rhvoice]
First create TTS object:
from rhvoice_wrapper import TTS
tts = TTS(threads=1)
You may set options when creating or through variable environments (UPPER REGISTER). Options override variable environments. To set the default value use None
:
1
, created one thread object, if more running in multiprocessing mode and create a lot of processes. Default 1
.True
engines run in multiprocessing mode, if False
in threads mode.
Default False
if threads == 1, else True
.
Threads mode and threads > 1 causes a segmentation faults or may return corrupted datalibRHVoice.so
in Linux, libRHVoice.dylib
in macOS and RHVoice.dll
in Windows./usr/local/share/RHVoice
./usr/local/etc/RHVoice
.[]
.lame
, optional. Lame must be present for mp3
support. Default lame
.opusenc
, optional. File must be present for opus
support. Default opusenc
.flac
, optional. File must be present for flac
support. Default flac
.True
don't info output. Default False
.say
will return one big chunk, formats other than wav
and pcm
will be generated much slower. Default True
.Start synthesis generator and get audio data, chunk by chunk:
def generator_audio(text, voice='anna', format_='wav', buff=4096, sets=None):
with tts.say(text, voice, format_, buff, sets) as gen:
for chunk in gen:
yield chunk
Or get all audio data in one big chunk:
data = tts.get('Hello world!', format_='wav')
print('data size: ', len(data), ' bytes')
subprocess.check_output(['aplay', '-q'], input=data)
Or just save to file:
tts.to_file(filename='esperanto.ogg', text='Saluton mondo', voice='spomenka', format_='opus', sets=None)
format_
is output audio format. Must be present in tts.formats
.
voice
is a voice of speaker. Must be present in tts.voice_profiles
.
voice='Voice', sets=None
equal voice=None, sets={'voice_profile': 'Voice'}
, voice
more priority.
sets
may set as dict containing synthesis parameters as in set_params.
This parameters only work for current phrase. Default None
.
If buff
equal None or 0
, for pcm and wav chunks return as is (probably little faster).
For others used default chunk size (4 KiB).
If text
iterable object, all its fragments will processing successively.
This is a good method for processing incredibly large texts.
Remember, the generator cannot be transferred to another process. Example:
def _text():
with open('wery_large_book.txt') as fp:
text = fp.read(5000)
while text:
yield text
text = fp.read(5000)
def generator_audio():
with tts.say(_text()) as gen:
for chunk in gen:
yield chunk
Changes voice synthesizer settings:
tts.set_params(**kwargs)
Allow: voice_profile
, absolute_rate
, absolute_pitch
, absolute_volume
, relative_rate
, relative_pitch
, relative_volume
, punctuation_mode
, punctuation_list
, capitals_mode
, flags
. See RHVoice documentation for details.
Return True
if change, else False
.
Get voice synthesizer settings:
tts.get_params(param=None)
If param is None
return all settings in dict
, else parameter value by name. If parameter not found return None
.
Join thread or processes. Don't use object after join:
tts.join()
TTS.formats
: List of supported formats, pcm
and wav
always present.TTS.thread_count
: Number of synthesis threads.TTS.process
: If True
, TTS running in multiprocessing mode.TTS.voices
: List of supported voices.TTS.voice_profiles
: List of supported voice profiles.TTS.voices_info
: Dictionary of supported voices with voices information. TTS.api_version
: Supported RHVoice library version.TTS.lib_version
: RHVoice library version. If not in rhvoice_wrapper.rhvoice_proxy.SUPPORT
, may incorrect work.TTS.cmd
: Dictionary of external calls, as it is.