1
0
Fork 0
forked from len0rd/rockbox

rbutil: commit FS#9983 by Delyan Kratunov. It adds support for the Festival TTS on Linux.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@20559 a1c6a512-1295-4272-9138-f99709370657
This commit is contained in:
Dominik Wenger 2009-03-27 19:18:14 +00:00
parent 75a5b9321b
commit 8f2aaaf79d
9 changed files with 540 additions and 31 deletions

View file

@ -8,3 +8,4 @@ Tomer Shalev
Yoshihisa Uchida
Alexander Spyridakis
Rui Araújo
Delyan Kratunov

View file

@ -171,6 +171,7 @@ FORMS += rbutilqtfrm.ui \
encexescfgfrm.ui \
ttsexescfgfrm.ui \
sapicfgfrm.ui \
ttsfestivalcfgform.ui \
createvoicefrm.ui \
sysinfofrm.ui

View file

@ -34,7 +34,7 @@ bool TalkFileCreator::createTalkFiles(ProgressloggerInterface* logger)
QMultiMap<QString,QString> fileList;
QMultiMap<QString,QString> dirList;
QStringList toSpeakList;
QStringList toSpeakList, voicedEntries, encodedEntries;
QString errStr;
m_logger->addItem(tr("Starting Talk file generation"),LOGINFO);
@ -103,20 +103,19 @@ bool TalkFileCreator::createTalkFiles(ProgressloggerInterface* logger)
}
}
// Voice entryies
// Voice entries
m_logger->addItem(tr("Voicing entries..."),LOGINFO);
if(voiceList(toSpeakList,&errStr) == false)
TTSStatus voiceStatus= voiceList(toSpeakList,voicedEntries);
if(voiceStatus == FatalError)
{
m_logger->addItem(errStr,LOGERROR);
doAbort(toSpeakList);
return false;
}
// Encoding Entries
m_logger->addItem(tr("Encoding files..."),LOGINFO);
if(encodeList(toSpeakList,&errStr) == false)
if(encodeList(voicedEntries,encodedEntries) == false)
{
m_logger->addItem(errStr,LOGERROR);
doAbort(toSpeakList);
return false;
}
@ -247,29 +246,45 @@ bool TalkFileCreator::createDirAndFileMaps(QDir startDir,QMultiMap<QString,QStri
//! \param toSpeak QStringList with the Entries to voice.
//! \param errString pointer to where the Error cause is written
//! \returns true on success, false on error or user abort
bool TalkFileCreator::voiceList(QStringList toSpeak,QString* errString)
TTSStatus TalkFileCreator::voiceList(QStringList toSpeak,QStringList& voicedEntries)
{
resetProgress(toSpeak.size());
QStringList errors;
bool warnings = false;
for(int i=0; i < toSpeak.size(); i++)
{
if(m_abort)
{
*errString = tr("Talk file creation aborted");
return false;
m_logger->addItem(tr("Talk file creation aborted"), LOGERROR);
return FatalError;
}
QString filename = QDir::tempPath()+ "/"+ toSpeak[i] + ".wav";
if(!m_tts->voice(toSpeak[i],filename))
QString error;
TTSStatus status = m_tts->voice(toSpeak[i],filename, &error);
if(status == Warning)
{
*errString =tr("Voicing of %s failed").arg(toSpeak[i]);
return false;
warnings = true;
m_logger->addItem(tr("Voicing of %1 failed: %2").arg(toSpeak[i]).arg(error),
LOGWARNING);
}
else if (status == FatalError)
{
m_logger->addItem(tr("Voicing of %1 failed: %2").arg(toSpeak[i]).arg(error),
LOGERROR);
return FatalError;
}
else
voicedEntries.append(toSpeak[i]);
m_logger->setProgressValue(++m_progress);
QCoreApplication::processEvents();
}
return true;
if(warnings)
return Warning;
else
return NoError;
}
@ -279,14 +294,14 @@ bool TalkFileCreator::voiceList(QStringList toSpeak,QString* errString)
//! \param toSpeak QStringList with the Entries to encode.
//! \param errString pointer to where the Error cause is written
//! \returns true on success, false on error or user abort
bool TalkFileCreator::encodeList(QStringList toEncode,QString* errString)
bool TalkFileCreator::encodeList(QStringList toEncode,QStringList& encodedEntries)
{
resetProgress(toEncode.size());
for(int i=0; i < toEncode.size(); i++)
{
if(m_abort)
{
*errString = tr("Talk file creation aborted");
m_logger->addItem(tr("Talk file creation aborted"), LOGERROR);
return false;
}
@ -295,9 +310,10 @@ bool TalkFileCreator::encodeList(QStringList toEncode,QString* errString)
if(!m_enc->encode(wavfilename,filename))
{
*errString =tr("Encoding of %1 failed").arg(filename);
m_logger->addItem(tr("Encoding of %1 failed").arg(filename), LOGERROR);
return false;
}
encodedEntries.append(toEncode[i]);
m_logger->setProgressValue(++m_progress);
QCoreApplication::processEvents();
}
@ -327,6 +343,10 @@ bool TalkFileCreator::copyTalkDirFiles(QMultiMap<QString,QString> dirMap,QString
}
QString source = QDir::tempPath()+ "/"+ it.value() + ".talk";
if(!QFileInfo(source).exists())
continue; // this file was skipped in one of the previous steps
QString target = it.key() + "/" + "_dirname.talk";
// remove target if it exists, and if we should overwrite it
@ -383,6 +403,9 @@ bool TalkFileCreator::copyTalkFileFiles(QMultiMap<QString,QString> fileMap,QStri
else
source = QDir::tempPath()+ "/"+ it.value() + ".talk";
if(!QFileInfo(source).exists())
continue; // this file was skipped in one of the previous steps
// remove target if it exists, and if we should overwrite it
if(m_overwriteTalk && QFile::exists(target))
QFile::remove(target);

View file

@ -58,8 +58,8 @@ private:
void doAbort(QStringList cleanupList);
void resetProgress(int max);
bool createDirAndFileMaps(QDir startDir,QMultiMap<QString,QString> *dirMap,QMultiMap<QString,QString> *fileMap);
bool voiceList(QStringList toSpeak,QString* errString);
bool encodeList(QStringList toEncode,QString* errString);
TTSStatus voiceList(QStringList toSpeak,QStringList& voicedEntries);
bool encodeList(QStringList toEncode,QStringList& encodedEntries);
bool copyTalkDirFiles(QMultiMap<QString,QString> dirMap,QString* errString);
bool copyTalkFileFiles(QMultiMap<QString,QString> fileMap,QString* errString);

View file

@ -33,7 +33,9 @@ void TTSBase::initTTSList()
#if defined(Q_OS_WIN)
ttsList["sapi"] = "Sapi TTS Engine";
#endif
#if defined(Q_OS_LINUX)
ttsList["festival"] = "Festival TTS Engine";
#endif
}
// function to get a specific encoder
@ -44,6 +46,7 @@ TTSBase* TTSBase::getTTS(QString ttsName)
return ttsCache.value(ttsName);
TTSBase* tts;
#if defined(Q_OS_WIN)
if(ttsName == "sapi")
{
tts = new TTSSapi();
@ -51,6 +54,17 @@ TTSBase* TTSBase::getTTS(QString ttsName)
return tts;
}
else
#endif
#if defined(Q_OS_LINUX)
if (ttsName == "festival")
{
tts = new TTSFestival();
ttsCache[ttsName] = tts;
return tts;
}
else
#endif
if (true) // fix for OS other than WIN or LINUX
{
tts = new TTSExes(ttsName);
ttsCache[ttsName] = tts;
@ -92,7 +106,7 @@ TTSExes::TTSExes(QString name) : TTSBase()
m_name = name;
m_TemplateMap["espeak"] = "\"%exe\" %options -w \"%wavfile\" \"%text\"";
m_TemplateMap["flite"] = "\"%exe\" %options -o \"%wavfile\" \"%text\"";
m_TemplateMap["flite"] = "\"%exe\" %options -o \"%wavfile\" -t \"%text\"";
m_TemplateMap["swift"] = "\"%exe\" %options -o \"%wavfile\" \"%text\"";
}
@ -153,8 +167,9 @@ bool TTSExes::start(QString *errStr)
}
}
bool TTSExes::voice(QString text,QString wavfile)
TTSStatus TTSExes::voice(QString text,QString wavfile, QString *errStr)
{
(void) errStr;
QString execstring = m_TTSTemplate;
execstring.replace("%exe",m_TTSexec);
@ -163,7 +178,7 @@ bool TTSExes::voice(QString text,QString wavfile)
execstring.replace("%text",text);
//qDebug() << "voicing" << execstring;
QProcess::execute(execstring);
return true;
return NoError;
}
@ -304,15 +319,16 @@ QStringList TTSSapi::getVoiceList(QString language)
bool TTSSapi::voice(QString text,QString wavfile)
TTSStatus TTSSapi::voice(QString text,QString wavfile, QString *errStr)
{
(void) errStr;
QString query = "SPEAK\t"+wavfile+"\t"+text+"\r\n";
qDebug() << "voicing" << query;
*voicestream << query;
*voicestream << "SYNC\tbla\r\n";
voicestream->flush();
voicescript->waitForReadyRead();
return true;
return NoError;
}
bool TTSSapi::stop()
@ -349,5 +365,254 @@ bool TTSSapi::configOk()
return false;
return true;
}
/**********************************************************************
* TSSFestival - client-server wrapper
**********************************************************************/
TTSFestival::~TTSFestival()
{
stop();
}
void TTSFestival::startServer()
{
if(!configOk())
return;
QStringList paths = settings->ttsPath("festival").split(":");
serverProcess.start(QString("%1 --server").arg(paths[0]));
serverProcess.waitForStarted();
queryServer("(getpid)");
if(serverProcess.state() == QProcess::Running)
qDebug() << "Festival is up and running";
else
qDebug() << "Festival failed to start";
}
void TTSFestival::ensureServerRunning()
{
if(serverProcess.state() != QProcess::Running)
{
// least common denominator for all the server startup code paths
QProgressDialog progressDialog(tr(""), tr(""), 0, 0);
progressDialog.setWindowTitle(tr("Starting festival"));
progressDialog.setModal(true);
progressDialog.setLabel(0);
progressDialog.setCancelButton(0);
progressDialog.show();
QApplication::processEvents(); // actually show the dialog
startServer();
}
}
bool TTSFestival::start(QString* errStr)
{
(void) errStr;
ensureServerRunning();
if (!settings->ttsVoice("festival").isEmpty())
queryServer(QString("(voice.select '%1)").arg(settings->ttsVoice("festival")));
return true;
}
bool TTSFestival::stop()
{
serverProcess.terminate();
serverProcess.kill();
return true;
}
TTSStatus TTSFestival::voice(QString text, QString wavfile, QString* errStr)
{
qDebug() << text << "->" << wavfile;
QStringList paths = settings->ttsPath("festival").split(":");
QString cmd = QString("%1 --server localhost --otype riff --ttw --withlisp --output \"%2\" - ").arg(paths[1]).arg(wavfile);
qDebug() << cmd;
QProcess clientProcess;
clientProcess.start(cmd);
clientProcess.write(QString("%1.\n").arg(text).toAscii());
clientProcess.waitForBytesWritten();
clientProcess.closeWriteChannel();
clientProcess.waitForReadyRead();
QString response = clientProcess.readAll();
response = response.trimmed();
if(!response.contains("Utterance"))
{
qDebug() << "Could not voice string: " << response;
*errStr = tr("engine could not voice string");
return Warning;
/* do not stop the voicing process because of a single string
TODO: needs proper settings */
}
clientProcess.closeReadChannel(QProcess::StandardError);
clientProcess.closeReadChannel(QProcess::StandardOutput);
clientProcess.terminate();
clientProcess.kill();
return NoError;
}
bool TTSFestival::configOk()
{
QStringList paths = settings->ttsPath("festival").split(":");
if(paths.size() != 2)
return false;
bool ret = QFileInfo(paths[0]).isExecutable() &&
QFileInfo(paths[1]).isExecutable();
if(settings->ttsVoice("festival").size() > 0 && voices.size() > 0)
ret = ret && (voices.indexOf(settings->ttsVoice("festival")) != -1);
return ret;
}
void TTSFestival::showCfg()
{
#ifndef CONSOLE
TTSFestivalGui gui(this);
#endif
gui.setCfg(settings);
gui.showCfg();
}
QStringList TTSFestival::getVoiceList()
{
if(!configOk())
return QStringList();
if(voices.size() > 0)
{
qDebug() << "Using voice cache";
return voices;
}
QString response = queryServer("(voice.list)");
// get the 2nd line. It should be (<voice_name>, <voice_name>)
response = response.mid(response.indexOf('\n') + 1, -1);
response = response.left(response.indexOf('\n')).trimmed();
voices = response.mid(1, response.size()-2).split(' ');
voices.sort();
if (voices.size() == 1 && voices[0].size() == 0)
voices.removeAt(0);
if (voices.size() > 0)
qDebug() << "Voices: " << voices;
else
qDebug() << "No voices.";
return voices;
}
QString TTSFestival::getVoiceInfo(QString voice)
{
if(!configOk())
return "";
if(!getVoiceList().contains(voice))
return "";
if(voiceDescriptions.contains(voice))
return voiceDescriptions[voice];
QString response = queryServer(QString("(voice.description '%1)").arg(voice), 3000);
if (response == "")
{
voiceDescriptions[voice]=tr("No description available");
}
else
{
response = response.remove(QRegExp("(description \"*\")", Qt::CaseInsensitive, QRegExp::Wildcard));
qDebug() << "voiceInfo w/o descr: " << response;
response = response.remove(')');
QStringList responseLines = response.split('(', QString::SkipEmptyParts);
responseLines.removeAt(0); // the voice name itself
QString description;
foreach(QString line, responseLines)
{
line = line.remove('(');
line = line.simplified();
line[0] = line[0].toUpper(); // capitalize the key
int firstSpace = line.indexOf(' ');
if (firstSpace > 0)
{
line = line.insert(firstSpace, ':'); // add a colon between the key and the value
line[firstSpace+2] = line[firstSpace+2].toUpper(); // capitalize the value
}
description += line + "\n";
}
voiceDescriptions[voice] = description.trimmed();
}
return voiceDescriptions[voice];
}
QString TTSFestival::queryServer(QString query, int timeout)
{
if(!configOk())
return "";
ensureServerRunning();
qDebug() << "queryServer with " << query;
QString response;
QDateTime endTime;
if(timeout > 0)
endTime = QDateTime::currentDateTime().addMSecs(timeout);
/* Festival is *extremely* unreliable. Although at this
* point we are sure that SIOD is accepting commands,
* we might end up with an empty response. Hence, the loop.
*/
while(true)
{
QApplication::processEvents(QEventLoop::AllEvents, 50);
QTcpSocket socket;
socket.connectToHost("localhost", 1314);
socket.waitForConnected();
if(socket.state() == QAbstractSocket::ConnectedState)
{
socket.write(QString("%1\n").arg(query).toAscii());
socket.waitForBytesWritten();
socket.waitForReadyRead();
response = socket.readAll().trimmed();
if (response != "LP" && response != "")
break;
}
socket.abort();
socket.disconnectFromHost();
if(timeout > 0 && QDateTime::currentDateTime() >= endTime)
return "";
/* make sure we wait a little as we don't want to flood the server with requests */
QDateTime tmpEndTime = QDateTime::currentDateTime().addMSecs(500);
while(QDateTime::currentDateTime() < tmpEndTime)
QApplication::processEvents(QEventLoop::AllEvents);
}
if(response == "nil")
return "";
QStringList lines = response.split('\n');
if(lines.size() > 2)
{
lines.removeFirst();
lines.removeLast();
}
else
qDebug() << "Response too short: " << response;
return lines.join("\n");
}

View file

@ -23,9 +23,13 @@
#ifndef TTS_H
#define TTS_H
#include "rbsettings.h"
#include <QtCore>
#include <QProcess>
#include <QProgressDialog>
#include <QDateTime>
#include <QRegExp>
#include <QTcpSocket>
#ifndef CONSOLE
#include "ttsgui.h"
@ -33,14 +37,18 @@
#include "ttsguicli.h"
#endif
enum TTSStatus{ FatalError, NoError, Warning };
class TTSSapi;
#if defined(Q_OS_LINUX)
class TTSFestival;
#endif
class TTSBase : public QObject
{
Q_OBJECT
public:
TTSBase();
virtual bool voice(QString text,QString wavfile)
{ (void)text; (void)wavfile; return false; }
virtual TTSStatus voice(QString text,QString wavfile, QString* errStr)
{ (void) text; (void) wavfile; (void) errStr; return FatalError;}
virtual bool start(QString *errStr) { (void)errStr; return false; }
virtual bool stop() { return false; }
virtual void showCfg(){}
@ -72,7 +80,7 @@ class TTSSapi : public TTSBase
Q_OBJECT
public:
TTSSapi();
virtual bool voice(QString text,QString wavfile);
virtual TTSStatus voice(QString text,QString wavfile, QString *errStr);
virtual bool start(QString *errStr);
virtual bool stop();
virtual void showCfg();
@ -99,7 +107,7 @@ class TTSExes : public TTSBase
Q_OBJECT
public:
TTSExes(QString name);
virtual bool voice(QString text,QString wavfile);
virtual TTSStatus voice(QString text,QString wavfile, QString *errStr);
virtual bool start(QString *errStr);
virtual bool stop() {return true;}
virtual void showCfg();
@ -115,4 +123,26 @@ class TTSExes : public TTSBase
QMap<QString,QString> m_TemplateMap;
};
class TTSFestival : public TTSBase
{
Q_OBJECT
public:
~TTSFestival();
virtual bool configOk();
virtual bool start(QString *errStr);
virtual bool stop();
virtual void showCfg();
virtual TTSStatus voice(QString text,QString wavfile, QString *errStr);
QStringList getVoiceList();
QString getVoiceInfo(QString voice);
private:
inline void startServer();
inline void ensureServerRunning();
QString queryServer(QString query, int timeout = -1);
QProcess serverProcess;
QStringList voices;
QMap<QString, QString> voiceDescriptions;
};
#endif

View file

@ -181,3 +181,159 @@ void TTSExesGui::browse()
}
}
TTSFestivalGui::TTSFestivalGui(TTSFestival* api, QDialog* parent) :
QDialog(parent), festival(api)
{
ui.setupUi(this);
this->setModal(true);
this->setDisabled(true);
this->show();
connect(ui.clientButton, SIGNAL(clicked()), this, SLOT(onBrowseClient()));
connect(ui.serverButton, SIGNAL(clicked()), this, SLOT(onBrowseServer()));
connect(ui.refreshButton, SIGNAL(clicked()), this, SLOT(onRefreshButton()));
connect(ui.voicesBox, SIGNAL(activated(QString)), this, SLOT(updateDescription(QString)));
connect(ui.showDescriptionCheckbox, SIGNAL(stateChanged(int)), this, SLOT(onShowDescription(int)));
}
void TTSFestivalGui::showCfg()
{
qDebug() << "show\tpaths: " << settings->ttsPath("festival") << "\n"
<< "\tvoice: " << settings->ttsVoice("festival");
// will populate the voices if the paths are correct,
// otherwise, it will require the user to press Refresh
updateVoices();
// try to get config from settings
QStringList paths = settings->ttsPath("festival").split(":");
if(paths.size() == 2)
{
ui.serverPath->setText(paths[0]);
ui.clientPath->setText(paths[1]);
}
this->setEnabled(true);
this->exec();
}
void TTSFestivalGui::accept(void)
{
//save settings in user config
QString newPath = QString("%1:%2").arg(ui.serverPath->text().trimmed()).arg(ui.clientPath->text().trimmed());
qDebug() << "set\tpaths: " << newPath << "\n\tvoice: " << ui.voicesBox->currentText();
settings->setTTSPath("festival", newPath);
settings->setTTSVoice("festival", ui.voicesBox->currentText());
settings->sync();
this->done(0);
}
void TTSFestivalGui::reject(void)
{
this->done(0);
}
void TTSFestivalGui::onBrowseClient()
{
BrowseDirtree browser(this);
browser.setFilter(QDir::Dirs | QDir::Files | QDir::NoDotAndDotDot);
QFileInfo currentPath(ui.clientPath->text().trimmed());
if(currentPath.isDir())
{
browser.setDir(ui.clientPath->text());
}
else if (currentPath.isFile())
{
browser.setDir(currentPath.dir().absolutePath());
}
if(browser.exec() == QDialog::Accepted)
{
qDebug() << browser.getSelected();
QString exe = browser.getSelected();
if(!QFileInfo(exe).isExecutable())
return;
ui.clientPath->setText(exe);
}
}
void TTSFestivalGui::onBrowseServer()
{
BrowseDirtree browser(this);
browser.setFilter(QDir::Dirs | QDir::Files | QDir::NoDotAndDotDot);
QFileInfo currentPath(ui.serverPath->text().trimmed());
if(currentPath.isDir())
{
browser.setDir(ui.serverPath->text());
}
else if (currentPath.isFile())
{
browser.setDir(currentPath.dir().absolutePath());
}
if(browser.exec() == QDialog::Accepted)
{
qDebug() << browser.getSelected();
QString exe = browser.getSelected();
if(!QFileInfo(exe).isExecutable())
return;
ui.serverPath->setText(exe);
}
}
void TTSFestivalGui::onRefreshButton()
{
/* Temporarily commit the settings so that we get the new path when we check for voices */
QString newPath = QString("%1:%2").arg(ui.serverPath->text().trimmed()).arg(ui.clientPath->text().trimmed());
QString oldPath = settings->ttsPath("festival");
qDebug() << "new path: " << newPath << "\n" << "old path: " << oldPath << "\nuse new: " << (newPath != oldPath);
if(newPath != oldPath)
{
qDebug() << "Using new paths for getVoiceList";
settings->setTTSPath("festival", newPath);
settings->sync();
}
updateVoices();
if(newPath != oldPath)
{
settings->setTTSPath("festival", oldPath);
settings->sync();
}
}
void TTSFestivalGui::onShowDescription(int state)
{
if(state == Qt::Unchecked)
ui.descriptionLabel->setText("");
else
updateDescription(ui.voicesBox->currentText());
}
void TTSFestivalGui::updateVoices()
{
ui.voicesBox->clear();
ui.voicesBox->addItem(tr("Loading.."));
QStringList voiceList = festival->getVoiceList();
ui.voicesBox->clear();
ui.voicesBox->addItems(voiceList);
ui.voicesBox->setCurrentIndex(ui.voicesBox->findText(settings->ttsVoice("festival")));
updateDescription(settings->ttsVoice("festival"));
}
void TTSFestivalGui::updateDescription(QString value)
{
if(ui.showDescriptionCheckbox->checkState() == Qt::Checked)
{
ui.descriptionLabel->setText(tr("Querying festival"));
ui.descriptionLabel->setText(festival->getVoiceInfo(value));
}
}

View file

@ -26,9 +26,11 @@
#include "ui_ttsexescfgfrm.h"
#include "ui_sapicfgfrm.h"
#include "ui_ttsfestivalcfgform.h"
class RbSettings;
class TTSSapi;
class TTSFestival;
class TTSSapiGui : public QDialog
{
@ -71,4 +73,32 @@ private:
QString m_name;
};
class TTSFestivalGui : public QDialog
{
Q_OBJECT
public:
TTSFestivalGui(TTSFestival* festival, QDialog* parent = NULL);
void showCfg();
void setCfg(RbSettings* sett){settings = sett;}
public slots:
virtual void accept(void);
virtual void reject(void);
//virtual void reset(void);
void onRefreshButton();
void onShowDescription(int state);
void onBrowseServer();
void onBrowseClient();
private:
Ui::TTSFestivalCfgFrm ui;
RbSettings* settings;
TTSFestival* festival;
void updateVoices();
private slots:
void updateDescription(QString value);
};
#endif

View file

@ -230,7 +230,10 @@ void VoiceFileCreator::downloadDone(bool error)
m_logger->addItem(tr("creating ")+toSpeak,LOGINFO);
QCoreApplication::processEvents();
m_tts->voice(toSpeak,wavname); // generate wav
// TODO: add support for aborting the operation
QString errStr;
m_tts->voice(toSpeak,wavname, &errStr); // generate wav
}
// todo strip