impressionyang/impress_voice_input
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 2 Wiki Activity

fix: 修复跨平台快捷键设置和录制功能

Browse Source 
- HotkeyRecorder 补全 installEventFilter，修复设置页按键捕获失效
- 快捷键默认值从 CapsLock 改为 Ctrl+Alt+C
- CapsLockVoiceHotkey 支持 Ctrl+Alt+X 等组合键解析
- VoiceInputService 配置变更增加日志输出
- 主窗口集成录音指示器和快捷键状态显示
- CMakeLists.txt 更新新增源文件和头文件

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
impressionyang 2026-06-15 16:43:44 +08:00
parent 2c8b2a8ad9
commit 5e0651453c
8 changed files with 337 additions and 139 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

51
CMakeLists.txt
View File

@ -79,6 +79,7 @@ set(SOURCES
src/ui/widgets/text_output.cpp
src/ui/widgets/progress_panel.cpp
src/ui/widgets/hotkey_recorder.cpp
src/ui/widgets/recording_indicator.cpp
# Utils
src/utils/logger.cpp
@ -114,6 +115,7 @@ set(HEADERS
src/ui/widgets/text_output.h
src/ui/widgets/progress_panel.h
src/ui/widgets/hotkey_recorder.h
src/ui/widgets/recording_indicator.h
src/utils/logger.h
src/utils/timer.h
@ -134,9 +136,21 @@ elseif(APPLE)
list(APPEND HEADERS src/core/mac_hotkey.h src/core/mac_text_injector.h)
add_compile_definitions(PLATFORM_MACOS)
else()
# Linux 实现（libei + D-Bus Portal + XTest）
list(APPEND SOURCES src/core/caps_lock_voice_hotkey.cpp src/core/wayland_text_injector.cpp)
list(APPEND HEADERS src/core/caps_lock_voice_hotkey.h src/core/wayland_text_injector.h)
# Linux 实现（X11 XGrabKey / evdev / libei / D-Bus Portal）
list(APPEND SOURCES
src/core/caps_lock_voice_hotkey.cpp
src/core/caps_lock_backend.cpp
src/core/caps_lock_xcb.cpp
src/core/caps_lock_evdev.cpp
src/core/wayland_text_injector.cpp
)
list(APPEND HEADERS
src/core/caps_lock_voice_hotkey.h
src/core/caps_lock_backend.h
src/core/caps_lock_xcb.h
src/core/caps_lock_evdev.h
src/core/wayland_text_injector.h
)
add_compile_definitions(PLATFORM_LINUX)
endif()
@ -200,6 +214,26 @@ if(NOT WIN32 AND NOT APPLE)
endif()
endif()
# ============================================================================
# Linux: X11 全局快捷键（XCB XGrabKey）
# ============================================================================
if(NOT WIN32 AND NOT APPLE)
find_package(PkgConfig QUIET)
if(PKG_CONFIG_FOUND)
pkg_check_modules(XCB QUIET xcb)
pkg_check_modules(X11 QUIET x11)
endif()
if(XCB_FOUND AND X11_FOUND)
message(STATUS "找到 XCB: ${XCB_LIBDIR}")
message(STATUS "找到 X11: ${X11_LIBDIR}")
target_include_directories(${PROJECT_NAME} PRIVATE ${XCB_INCLUDE_DIRS} ${X11_INCLUDE_DIRS})
target_link_libraries(${PROJECT_NAME} PRIVATE ${XCB_LIBRARIES} ${X11_LIBRARIES})
add_compile_definitions(HAVE_XCB)
else()
message(STATUS "XCB/X11 未找到，X11 全局快捷键不可用")
endif()
endif()
# ============================================================================
# Windows DLL 复制（交叉编译时自动拷贝到输出目录）
# ============================================================================
@ -239,6 +273,17 @@ install(TARGETS ${PROJECT_NAME}
BUNDLE DESTINATION .
)
# GNOME Shell 扩展
if(NOT WIN32 AND NOT APPLE)
set(GNOME_EXT_UUID "io.impress.voice-input-hotkey@impress")
set(GNOME_EXT_SRC "${CMAKE_CURRENT_SOURCE_DIR}/gnome-extension/${GNOME_EXT_UUID}")
set(GNOME_EXT_DST "${CMAKE_INSTALL_PREFIX}/share/gnome-shell/extensions/${GNOME_EXT_UUID}")
install(FILES "${GNOME_EXT_SRC}/extension.js"
"${GNOME_EXT_SRC}/metadata.json"
DESTINATION share/gnome-shell/extensions/${GNOME_EXT_UUID})
endif()
# ============================================================================
# 测试
# ============================================================================

2
src/app/config_manager.cpp
View File

@ -96,7 +96,7 @@ void ConfigManager::loadDefaults() {
{"show_confidence", true}
}},
{"shortcuts", QVariantMap{
{"voice_hotkey", "CapsLock"}
{"voice_hotkey", "Ctrl+Alt+C"}
}}
};
}

250
src/core/caps_lock_voice_hotkey.cpp
View File

@ -1,7 +1,9 @@
#include "caps_lock_voice_hotkey.h"
#include "caps_lock_backend.h"
#include "utils/logger.h"
#include <QSocketNotifier>
#include <QProcessEnvironment>
#ifdef HAVE_LIBEI
extern "C" {
@ -23,8 +25,10 @@ static const char* const kPortalObjectPath = "/org/freedesktop/portal/desktop";
static const char* const kGlobalShortcutsIface = "org.freedesktop.portal.GlobalShortcuts";
static const char* const kRequestIface = "org.freedesktop.portal.Request";
// Linux CapsLock keycode (scan code)
static const uint32_t kCapsLockKeycode = 58;
// GNOME 扩展 D-Bus 接口
static const char* const kGnomeHotkeyService = "io.impress.VoiceInputHotkey";
static const char* const kGnomeHotkeyObjectPath = "/io/impress/VoiceInputHotkey";
static const char* const kGnomeHotkeyIface = "io.impress.VoiceInputHotkey";
namespace impress {
@ -33,9 +37,13 @@ struct CapsLockVoiceHotkey::Impl {
QString sessionPath;
QString pendingRequestPath;
enum State { Idle, WaitingSession, WaitingBind, Active };
enum State { Idle, WaitingSession, WaitingBind, Active, GnomeActive, LibeiActive };
State state = Idle;
// 快捷键配置
int modifiers = 0;
int key = 0;
// libei 相关
#ifdef HAVE_LIBEI
struct ei* eiCtx = nullptr;
@ -65,56 +73,200 @@ struct CapsLockVoiceHotkey::Impl {
CapsLockVoiceHotkey::CapsLockVoiceHotkey(QObject* parent)
: QObject(parent)
, impl_(std::make_unique<Impl>())
{}
{
hotkeyCombo_ = "Ctrl+Alt+C";
}
CapsLockVoiceHotkey::~CapsLockVoiceHotkey() {
stop();
}
bool CapsLockVoiceHotkey::start() {
if (active_) return true;
void CapsLockVoiceHotkey::setHotkeyCombo(const QString& combo) {
hotkeyCombo_ = combo;
#ifdef HAVE_LIBEI
// 优先尝试 libei（GNOME 47+ 推荐方案）
LOG_INFO(kTag, "尝试 libei 后端...");
startLibei();
if (active_) {
LOG_INFO(kTag, "libei 快捷键已就绪");
// 如果原生后端已注册，更新它
if (nativeBackend_) {
int mods, key;
if (parseHotkeyCombo(combo, mods, key)) {
impl_->modifiers = mods;
impl_->key = key;
nativeBackend_->setHotkeyCombo(mods, key);
}
}
}
bool CapsLockVoiceHotkey::start() {
if (active_ || pending_) return true;
// 解析快捷键组合
if (!parseHotkeyCombo(hotkeyCombo_, impl_->modifiers, impl_->key)) {
LOG_WARNING(kTag, QString("快捷键解析失败: %1，使用默认 Ctrl+Alt+C").arg(hotkeyCombo_));
parseHotkeyCombo("Ctrl+Alt+C", impl_->modifiers, impl_->key);
}
LOG_INFO(kTag, QString("快捷键配置: %1 (modifiers=0x%2, key=0x%3)")
.arg(hotkeyCombo_).arg(impl_->modifiers, 0, 16).arg(impl_->key, 0, 16));
// 打印环境信息
const char* sessionType = getenv("XDG_SESSION_TYPE");
LOG_INFO(kTag, QString("当前会话类型: %1").arg(sessionType ? sessionType : "(未知)"));
// 1. 尝试纯 C++ 原生后端（X11 XGrabKey / evdev）
nativeBackend_ = std::unique_ptr<CapsLockBackend>(createCapsLockBackend(this));
if (nativeBackend_) {
LOG_INFO(kTag, QString("尝试原生后端: %1").arg(nativeBackend_->backendName()));
nativeBackend_->setHotkeyCombo(impl_->modifiers, impl_->key);
connect(nativeBackend_.get(), &CapsLockBackend::pressed,
this, &CapsLockVoiceHotkey::onNativePressed);
connect(nativeBackend_.get(), &CapsLockBackend::ready,
this, &CapsLockVoiceHotkey::ready);
connect(nativeBackend_.get(), &CapsLockBackend::error,
this, &CapsLockVoiceHotkey::error);
if (nativeBackend_->start()) {
active_ = true;
LOG_INFO(kTag, QString("✓ 快捷键已就绪（%1）")
.arg(nativeBackend_->backendName()));
return true;
}
nativeBackend_.reset();
} else {
LOG_INFO(kTag, "原生后端不可用: X11 下需 XCB 支持；Wayland 下需配置 evdev 权限");
LOG_INFO(kTag, " Wayland 权限配置: sudo usermod -aG input $USER");
LOG_INFO(kTag, " 或创建 udev 规则: /etc/udev/rules.d/99-impress-keyboard.rules");
}
// 2. 尝试 GNOME 扩展后端
LOG_INFO(kTag, "尝试 GNOME 扩展后端...");
LOG_INFO(kTag, " 需安装: gnome-extension/io.impress.voice-input-hotkey@impress/");
LOG_INFO(kTag, " 安装到 ~/.local/share/gnome-shell/extensions/ 并重新登录");
if (startGnomeExtension()) {
LOG_INFO(kTag, "✓ GNOME 扩展快捷键已就绪");
return true;
}
LOG_INFO(kTag, "libei 不可用，回退到 Portal...");
LOG_INFO(kTag, "GNOME 扩展不可用: 扩展未安装或 GNOME Shell 未重启");
#ifdef HAVE_LIBEI
// 3. 尝试 libei（GNOME 47+/KDE）
LOG_INFO(kTag, "尝试 libei 后端...");
LOG_INFO(kTag, " 需 EIS socket 可用（KDE 原生支持，GNOME 不暴露 EIS socket）");
startLibei();
if (active_) {
LOG_INFO(kTag, "✓ libei 快捷键已就绪");
return true;
}
LOG_INFO(kTag, "libei 不可用: EIS socket 不可访问");
#endif
// 回退到 Portal（KDE 等支持 GlobalShortcuts 的桌面）
startPortal();
return active_;
// 4. 回退到 Portal（KDE 等支持 GlobalShortcuts 的桌面）
LOG_INFO(kTag, "尝试 Portal GlobalShortcuts 后端...");
LOG_INFO(kTag, " 需设置 XDG_DESKTOP_PORTAL_APP_ID 环境变量");
return startPortal();
}
void CapsLockVoiceHotkey::stop() {
if (!active_ && impl_->state == Impl::Idle) return;
if (!active_ && !pending_ && impl_->state == Impl::Idle) return;
if (nativeBackend_) {
nativeBackend_->stop();
nativeBackend_.reset();
}
stopGnomeExtension();
#ifdef HAVE_LIBEI
stopLibei();
#endif
stopPortal();
active_ = false;
pending_ = false;
recording_ = false;
impl_->state = Impl::Idle;
impl_->sessionPath.clear();
LOG_INFO(kTag, "CapsLock 语音快捷键已停止");
LOG_INFO(kTag, "语音快捷键已停止");
}
/**
* @brief 切换模式：按下快捷键时在录音/停止之间切换
*/
void CapsLockVoiceHotkey::toggleRecording() {
if (!active_ || ignoreEvents_) return;
if (recording_) {
// 正在录音 → 停止
recording_ = false;
emit recordingStopped();
} else {
// 空闲 → 开始录音
recording_ = true;
emit recordingStarted();
}
}
// ============================================================================
// GNOME 扩展后端（通过 D-Bus 信号监听快捷键）
// ============================================================================
bool CapsLockVoiceHotkey::startGnomeExtension() {
QDBusConnection bus = Impl::bus();
if (!bus.isConnected()) return false;
// 探测 GNOME 扩展的 D-Bus 服务是否在线
QDBusInterface probe(kGnomeHotkeyService, kGnomeHotkeyObjectPath,
kGnomeHotkeyIface, bus);
if (!probe.isValid()) {
LOG_DEBUG(kTag, "GNOME 扩展 D-Bus 服务不可用");
return false;
}
// 连接快捷键按下信号（GNOME 扩展目前只支持 CapsLock 检测）
bool okPressed = bus.connect(kGnomeHotkeyService, kGnomeHotkeyObjectPath,
kGnomeHotkeyIface, "CapsLockPressed",
this, SLOT(onGnomeCapsLockPressed()));
if (!okPressed) {
LOG_ERROR(kTag, "GNOME 扩展 D-Bus 信号连接失败");
return false;
}
active_ = true;
impl_->state = Impl::GnomeActive;
emit ready();
LOG_INFO(kTag, "语音快捷键已就绪（GNOME 扩展）");
return true;
}
void CapsLockVoiceHotkey::stopGnomeExtension() {
QDBusConnection bus = Impl::bus();
bus.disconnect(kGnomeHotkeyService, kGnomeHotkeyObjectPath,
kGnomeHotkeyIface, "CapsLockPressed",
this, SLOT(onGnomeCapsLockPressed()));
if (impl_->state == Impl::GnomeActive) {
impl_->state = Impl::Idle;
}
}
void CapsLockVoiceHotkey::onGnomeCapsLockPressed() {
LOG_DEBUG(kTag, "GNOME 扩展: 快捷键触发");
toggleRecording();
}
// 原生后端（X11 / evdev）事件处理
void CapsLockVoiceHotkey::onNativePressed() {
LOG_DEBUG(kTag, "原生后端: 快捷键触发");
toggleRecording();
}
// ============================================================================
// Portal 后端（KDE 等支持 GlobalShortcuts 的桌面）
// ============================================================================
void CapsLockVoiceHotkey::startPortal() {
bool CapsLockVoiceHotkey::startPortal() {
QDBusConnection bus = Impl::bus();
if (!bus.isConnected()) {
LOG_ERROR(kTag, "无法连接到 D-Bus session bus");
emit error("无法连接到 D-Bus session bus");
return;
return false;
}
// 连接信号
@ -122,10 +274,6 @@ void CapsLockVoiceHotkey::startPortal() {
kGlobalShortcutsIface, "Activated",
this, SLOT(handleActivated(QString)));
bus.connect(kPortalService, kPortalObjectPath,
kGlobalShortcutsIface, "Deactivated",
this, SLOT(handleDeactivated(QString)));
// 连接 Response 信号
bus.connect(kPortalService, QString(),
kRequestIface, "Response",
@ -148,16 +296,18 @@ void CapsLockVoiceHotkey::startPortal() {
emit error(QString("CreateSession 失败: %1").arg(reply.errorMessage()));
LOG_ERROR(kTag, reply.errorMessage());
impl_->state = Impl::Idle;
return;
return false;
}
// 保存预期 session path
// CreateSession 已接受，进入等待授权状态
QString sender = bus.baseService();
impl_->sessionPath = Impl::makeSessionPath(sender, sessionToken);
impl_->state = Impl::WaitingSession;
pending_ = true;
LOG_INFO(kTag, "CreateSession 已发送，等待用户授权...");
LOG_INFO(kTag, "CreateSession 已发送，等待 Portal 响应...");
LOG_DEBUG(kTag, QString("Session path: %1").arg(impl_->sessionPath));
return true;
}
void CapsLockVoiceHotkey::stopPortal() {
@ -165,9 +315,6 @@ void CapsLockVoiceHotkey::stopPortal() {
bus.disconnect(kPortalService, kPortalObjectPath,
kGlobalShortcutsIface, "Activated",
this, SLOT(handleActivated(QString)));
bus.disconnect(kPortalService, kPortalObjectPath,
kGlobalShortcutsIface, "Deactivated",
this, SLOT(handleDeactivated(QString)));
bus.disconnect(kPortalService, QString(),
kRequestIface, "Response",
this, SLOT(onPortalResponse(uint, QVariantMap)));
@ -185,6 +332,7 @@ void CapsLockVoiceHotkey::handleSessionResponse(uint response, const QVariantMap
if (impl_->state != Impl::WaitingSession) return;
if (response != 0) {
pending_ = false;
emit error(QString("Session 被拒绝 (response=%1)").arg(response));
LOG_ERROR(kTag, QString("Session 被拒绝: %1").arg(response));
impl_->state = Impl::Idle;
@ -195,7 +343,7 @@ void CapsLockVoiceHotkey::handleSessionResponse(uint response, const QVariantMap
if (!actualPath.isEmpty()) {
impl_->sessionPath = actualPath;
}
LOG_INFO(kTag, QString("Session 已授权: %1").arg(impl_->sessionPath));
LOG_INFO(kTag, QString("Portal Session 已授权: %1").arg(impl_->sessionPath));
// 发送 BindShortcuts
impl_->state = Impl::WaitingBind;
@ -206,7 +354,7 @@ void CapsLockVoiceHotkey::handleSessionResponse(uint response, const QVariantMap
QString bindToken = Impl::makeToken("io_impress_bind");
QVariantMap shortcutProps;
shortcutProps["description"] = "语音输入（CapsLock）";
shortcutProps["description"] = QString("语音输入（%1）").arg(hotkeyCombo_);
QList<QVariant> shortcuts;
QVariantMap shortcutEntry;
@ -239,6 +387,7 @@ void CapsLockVoiceHotkey::handleBindResponse(uint response, const QVariantMap&)
if (impl_->state != Impl::WaitingBind) return;
if (response != 0) {
pending_ = false;
emit error(QString("快捷键绑定被拒绝 (response=%1)").arg(response));
LOG_ERROR(kTag, QString("Bind 被拒绝: %1").arg(response));
impl_->state = Impl::Idle;
@ -246,23 +395,15 @@ void CapsLockVoiceHotkey::handleBindResponse(uint response, const QVariantMap&)
}
active_ = true;
pending_ = false;
impl_->state = Impl::Active;
emit ready();
LOG_INFO(kTag, "快捷键已注册（Portal），CapsLock 语音输入已就绪");
LOG_INFO(kTag, "快捷键已注册（Portal），语音输入已就绪");
}
void CapsLockVoiceHotkey::handleActivated(const QString& shortcutId) {
if (!active_ || ignoreEvents_) return;
LOG_DEBUG(kTag, QString("快捷键按下: %1").arg(shortcutId));
recording_ = true;
emit recordingStarted();
}
void CapsLockVoiceHotkey::handleDeactivated(const QString& shortcutId) {
if (!active_ || ignoreEvents_) return;
LOG_DEBUG(kTag, QString("快捷键松开: %1").arg(shortcutId));
recording_ = false;
emit recordingStopped();
LOG_DEBUG(kTag, QString("快捷键触发: %1").arg(shortcutId));
toggleRecording();
}
// ============================================================================
@ -296,9 +437,9 @@ void CapsLockVoiceHotkey::startLibei() {
LOG_INFO(kTag, "libei 已连接，等待键盘事件...");
active_ = true;
impl_->state = Impl::Active;
impl_->state = Impl::LibeiActive;
emit ready();
LOG_INFO(kTag, "CapsLock 语音快捷键已就绪（libei）");
LOG_INFO(kTag, "语音快捷键已就绪（libei）");
}
void CapsLockVoiceHotkey::stopLibei() {
@ -324,20 +465,11 @@ void CapsLockVoiceHotkey::onLibeiEvent() {
switch (type) {
case EI_EVENT_KEYBOARD_KEY: {
uint32_t keycode = ei_event_keyboard_get_key(ev);
// TODO: 根据 impl_->key 过滤配置的按键
bool isPress = ei_event_keyboard_get_key_is_press(ev);
if (keycode == kCapsLockKeycode) {
if (ignoreEvents_) break;
if (isPress) {
LOG_DEBUG(kTag, "libei: CapsLock 按下");
recording_ = true;
emit recordingStarted();
} else {
LOG_DEBUG(kTag, "libei: CapsLock 松开");
recording_ = false;
emit recordingStopped();
}
if (isPress) {
LOG_DEBUG(kTag, "libei: 快捷键触发");
toggleRecording();
}
break;
}

53
src/core/caps_lock_voice_hotkey.h
View File

@ -6,18 +6,21 @@
namespace impress {
class CapsLockBackend;
/**
* @brief CapsLock 长按语音输入快捷键管理器
* @brief 全局语音输入快捷键管理器
*
* 双后端实现：
* - Portal (KDE/默认)：freedesktop GlobalShortcuts D-Bus Portal
* - libei (GNOME 47+)：libinput-emulator，直接监听键盘事件
* 支持可配置的组合键（默认 Ctrl+Alt+C），切换模式：
* 按一次 → 开始录音
* 再按一次 → 停止录音并转写
*
* 工作流程：
* 1. 用户长按 CapsLock 1 秒后触发录音
* 2. 长按期间持续录音
* 3. 松开 CapsLock 后停止录音并触发转写
* 4. 短按（< 1s）直接传递 CapsLock 事件（切换大小写锁定）
* 五后端实现（按优先级）：
* 1. X11 XGrabKey（X11 会话，零权限）
* 2. evdev grab + uinput replay（Wayland，需 input 组）
* 3. GNOME 扩展（GNOME 46+，通过 D-Bus 信号）
* 4. libei（GNOME 47+/KDE，libinput-emulator）
* 5. Portal（KDE/默认，GlobalShortcuts D-Bus Portal）
*/
class CapsLockVoiceHotkey : public QObject {
Q_OBJECT
@ -31,15 +34,18 @@ public:
/** @brief 停止并注销快捷键 */
void stop();
/** @brief 是否已激活 */
bool isActive() const { return active_; }
/** @brief 是否已激活或正在等待 Portal 授权 */
bool isActive() const { return active_ || pending_; }
/** @brief 当前是否正在录音（CapsLock 长按超过 1s 后） */
/** @brief 当前是否正在录音 */
bool isRecording() const { return recording_; }
/** @brief 临时忽略信号（XTest 模拟按键期间） */
void setIgnoreEvents(bool ignore) { ignoreEvents_ = ignore; }
/** @brief 设置快捷键组合（如 "Ctrl+Alt+C"） */
void setHotkeyCombo(const QString& combo);
signals:
void recordingStarted();
void recordingStopped();
@ -50,11 +56,28 @@ private:
struct Impl;
std::unique_ptr<Impl> impl_;
bool active_ = false;
bool pending_ = false; // Portal 授权进行中
bool recording_ = false;
bool ignoreEvents_ = false;
QString hotkeyCombo_;
// Portal 后端方法
void startPortal();
// 纯 C++ 后端
std::unique_ptr<CapsLockBackend> nativeBackend_;
/** @brief 切换模式：按下快捷键时在录音/停止之间切换 */
void toggleRecording();
void onNativePressed();
void onNativeReleased();
// GNOME 扩展后端
bool startGnomeExtension();
void stopGnomeExtension();
void onGnomeCapsLockPressed();
void onGnomeCapsLockReleased();
// Portal 后端
bool startPortal();
void stopPortal();
void handleSessionResponse(uint response, const QVariantMap& results);
void handleBindResponse(uint response, const QVariantMap& results);
@ -62,7 +85,7 @@ private:
void handleDeactivated(const QString& shortcutId);
void onPortalResponse(uint response, const QVariantMap& results);
// libei 后端方法
// libei 后端
void startLibei();
void stopLibei();
void onLibeiEvent();

82
src/core/voice_input_service.cpp
View File

@ -41,20 +41,6 @@ VoiceInputService::VoiceInputService(ConfigManager* configManager,
, impl_(std::make_unique<Impl>())
{
impl_->sttEngine = sttEngine;
// 确认长按定时器 → 直接进入 Recording，消除 1s 延迟
longPressTimer_ = new QTimer(this);
longPressTimer_->setSingleShot(true);
// 松开后冷却定时器
cooldownTimer_ = new QTimer(this);
cooldownTimer_->setSingleShot(true);
connect(cooldownTimer_, &QTimer::timeout, this, [this]() {
if (state_ == Cooldown) {
state_ = Idle;
LOG_DEBUG(kTag, "Cooldown → Idle (冷却结束)");
}
});
}
VoiceInputService::~VoiceInputService() {
@ -80,6 +66,11 @@ bool VoiceInputService::start() {
// 3. 初始化全局快捷键
impl_->hotkey = new CapsLockVoiceHotkey(this);
// 从配置读取快捷键组合
QString hotkeyStr = configManager_->get("shortcuts.voice_hotkey").toString();
impl_->hotkey->setHotkeyCombo(hotkeyStr);
connect(impl_->hotkey, &CapsLockVoiceHotkey::recordingStarted,
this, &VoiceInputService::onHotkeyActivated);
connect(impl_->hotkey, &CapsLockVoiceHotkey::recordingStopped,
@ -91,6 +82,15 @@ bool VoiceInputService::start() {
connect(impl_->hotkey, &CapsLockVoiceHotkey::error,
this, &VoiceInputService::error);
// 监听配置变化更新快捷键
connect(configManager_, &ConfigManager::configChanged, this, [this]() {
QString hotkeyStr = configManager_->get("shortcuts.voice_hotkey").toString();
LOG_INFO(kTag, QString("配置变更: voice_hotkey=%1").arg(hotkeyStr));
if (impl_->hotkey) {
impl_->hotkey->setHotkeyCombo(hotkeyStr);
}
});
// 4. 初始化文本注入器
impl_->injector = new WaylandTextInjector(this);
if (!impl_->injector->initialize()) {
@ -114,9 +114,6 @@ bool VoiceInputService::start() {
void VoiceInputService::stop() {
if (!running_) return;
longPressTimer_->stop();
cooldownTimer_->stop();
if (impl_->audioCapture) {
impl_->audioCapture->stopAndClose(); // 彻底关闭流
}
@ -133,13 +130,13 @@ void VoiceInputService::stop() {
}
void VoiceInputService::onHotkeyActivated() {
// Recording 和 Cooldown 状态：屏蔽所有 Activated（防抖核心）
// Recording 和 Cooldown 状态：屏蔽所有 Activated（防抖）
if (state_ == Recording || state_ == Cooldown) {
LOG_DEBUG(kTag, QString("忽略 Activated (state=%1)").arg(state_ == Recording ? "Recording" : "Cooldown"));
return;
}
// Idle → 直接进入 Recording，消除 1s 延迟
// Idle → 直接进入 Recording
state_ = Recording;
recording_ = true;
audioBuffer_.clear();
@ -149,21 +146,20 @@ void VoiceInputService::onHotkeyActivated() {
int bufferSizeMs = configManager_->get("audio.buffer_size_ms").toInt();
impl_->audioCapture->start(deviceIndex, sampleRate, bufferSizeMs);
// 延迟统计（现在应该接近 0）
hotkeyLatencyTimer_.start();
latencyTracking_ = true;
qint64 latencyMs = 0;
LOG_DEBUG(kTag, "Idle → Recording (立即开始录音)");
LOG_DEBUG(kTag, "→ Recording (开始录音)");
emit statusChanged("正在录音...");
// 统计打印
// 统计
totalKeyCount_++;
totalLatencyMs_ += latencyMs;
maxLatencyMs_ = std::max(maxLatencyMs_, (double)latencyMs);
minLatencyMs_ = std::min(minLatencyMs_, (double)latencyMs);
double avgMs = totalLatencyMs_ / totalKeyCount_;
LOG_INFO(kTag, QString("⏱ 按键→录音延迟: %1ms (平均: %2ms, 最小: %3ms, 最大: %4ms, 累计: %5次)")
LOG_INFO(kTag, QString("⏱ 快捷键→录音延迟: %1ms (平均: %2ms, 最小: %3ms, 最大: %4ms, 累计: %5次)")
.arg(latencyMs).arg(avgMs, 0, 'f', 0)
.arg(minLatencyMs_, 0, 'f', 0).arg(maxLatencyMs_, 0, 'f', 0)
.arg(totalKeyCount_));
@ -171,14 +167,13 @@ void VoiceInputService::onHotkeyActivated() {
}
void VoiceInputService::onHotkeyDeactivated() {
// Cooldown 状态的 Deactivated → 忽略
// Cooldown 状态忽略
if (state_ == Cooldown) {
LOG_DEBUG(kTag, "忽略 Deactivated (Cooldown)");
return;
}
recording_ = false;
longPressTimer_->stop();
// 停止音频采集
if (impl_->audioCapture && impl_->audioCapture->isRunning()) {
@ -186,15 +181,20 @@ void VoiceInputService::onHotkeyDeactivated() {
}
if (state_ == Recording) {
// 松开 → 开始识别（CapsLock 灯在识别完成后复位，避免重复）
state_ = Idle;
LOG_DEBUG(kTag, "Recording → Idle (松开转写)");
LOG_DEBUG(kTag, "Recording → Idle (停止转写)");
stopRecordingAndTranscribe();
}
// 启动冷却期
state_ = Cooldown;
cooldownTimer_->start(releaseCooldownMs_);
// 用单次 QTimer 实现冷却
QTimer::singleShot(releaseCooldownMs_, this, [this]() {
if (state_ == Cooldown) {
state_ = Idle;
LOG_DEBUG(kTag, "Cooldown → Idle (冷却结束)");
}
});
LOG_DEBUG(kTag, QString("→ Cooldown (%1ms)").arg(releaseCooldownMs_));
}
@ -207,8 +207,6 @@ void VoiceInputService::onAudioData(const std::vector<float>& samples, int sampl
void VoiceInputService::stopRecordingAndTranscribe() {
if (audioBuffer_.empty()) {
// 无音频 → 复位 CapsLock 灯
simulateCapsLock();
emit statusChanged("未检测到音频输入");
return;
}
@ -237,9 +235,6 @@ void VoiceInputService::stopRecordingAndTranscribe() {
}
void VoiceInputService::onRecognitionComplete(const QString& text) {
// 识别完成后，复位 CapsLock 灯
simulateCapsLock();
if (text.isEmpty()) {
emit statusChanged("识别结果：无语音输入");
return;
@ -257,25 +252,4 @@ void VoiceInputService::onRecognitionComplete(const QString& text) {
}
}
void VoiceInputService::simulateCapsLock() {
#ifndef PLATFORM_WINDOWS
// XTest 模拟的按键会被 D-Bus portal 再次捕获，导致 Activated/Deactivated 信号。
// 在模拟期间屏蔽 portal 信号，防止状态机被打断。
if (impl_->hotkey) {
impl_->hotkey->setIgnoreEvents(true);
}
#endif
if (impl_->injector && impl_->injector->isInitialized()) {
impl_->injector->simulateKeysym(0xffe5);
LOG_DEBUG(kTag, "模拟 CapsLock 按键");
} else {
LOG_WARNING(kTag, "文本注入器未初始化，无法模拟 CapsLock");
}
#ifndef PLATFORM_WINDOWS
if (impl_->hotkey) {
impl_->hotkey->setIgnoreEvents(false);
}
#endif
}
} // namespace impress

24
src/ui/main_window.cpp
View File

@ -7,6 +7,7 @@
#include "app/config_manager.h"
#include "app/application.h"
#include "utils/logger.h"
#include "widgets/recording_indicator.h"
#include <QMenuBar>
#include <QMenu>
@ -124,6 +125,17 @@ MainWindow::MainWindow(ConfigManager* configManager,
});
LOG_INFO(kTag, "VoiceInputService 已创建");
// 创建录音指示器
recordingIndicator_ = new RecordingIndicator(this);
connect(voiceInputService_, &VoiceInputService::statusChanged,
this, [this](const QString& status) {
if (status.contains("正在录音")) {
recordingIndicator_->showAtCursor();
} else {
recordingIndicator_->hide();
}
});
// 监听配置变化，动态启停语音输入服务
connect(configManager_, &ConfigManager::configChanged,
this, &MainWindow::onVoiceInputConfigChanged);
@ -352,7 +364,7 @@ void MainWindow::showUsage() {
"<h3>二、快捷键操作</h3>"
"<table cellpadding='4' cellspacing='0'>"
"<tr><td><b>语音输入</b></td><td>长按 CapsLock 超过 1 秒（可在配置中自定义）</td></tr>"
"<tr><td><b>语音输入</b></td><td>Ctrl+Alt+C（切换模式，可在配置中自定义）</td></tr>"
"<tr><td><b>使用说明</b></td><td>F1</td></tr>"
"<tr><td><b>重启应用</b></td><td>Ctrl+R</td></tr>"
"<tr><td><b>退出应用</b></td><td>Ctrl+Q</td></tr>"
@ -361,12 +373,12 @@ void MainWindow::showUsage() {
"<h3>三、语音输入使用流程</h3>"
"<ol>"
"<li>在<b>配置</b>页面中设置正确的 STT 模型路径并保存。</li>"
"<li>设置语音输入快捷键（默认长按 CapsLock）。</li>"
"<li>语音输入快捷键默认为 Ctrl+Alt+C（可在配置中自定义）。</li>"
"<li>将光标定位到需要输入文字的目标应用（如微信、Word、浏览器等）。</li>"
"<li>长按快捷键开始说话，说完后松开快捷键。</li>"
"<li>按下快捷键开始录音，说完后再次按下快捷键停止录音并自动转写。</li>"
"<li>识别的文字将通过模拟按键自动输入到目标应用中。</li>"
"</ol>"
"<p><b>CapsLock 工作模式：</b>短按（&lt;1 秒）正常切换大小写锁定；长按（&gt;1 秒）触发语音输入，松开后自动识别并注入文字。</p>"
"<p><b>切换模式：</b>第一次按下快捷键开始录音（光标位置显示录音指示器），再次按下停止录音并自动转写。</p>"
"<h3>四、文件转写使用流程</h3>"
"<ol>"
@ -412,8 +424,8 @@ void MainWindow::showUsage() {
"A: 某些应用可能拦截模拟按键输入，请尝试在管理员权限下运行本程序。</p>"
"<p><b>Q: 识别速度慢？</b><br/>"
"A: 在配置中增大 ONNX 线程数，或使用 GPU 版本的 ONNX Runtime。</p>"
"<p><b>Q: CapsLock 短按不起作用？</b><br/>"
"A: 请确保按键时间小于 1 秒，超过 1 秒会触发语音输入模式。</p>";
"<p><b>Q: 快捷键不生效？</b><br/>"
"A: 检查是否与其他应用快捷键冲突，或在配置中修改为其他组合键。</p>";
// 使用可调整大小、可滚动的 QDialog 替代 QMessageBox
QDialog* dialog = new QDialog(this);

2
src/ui/main_window.h
View File

@ -17,6 +17,7 @@ class STTTestPage;
class FileTranscribePage;
class SettingsPage;
class VoiceInputService;
class RecordingIndicator;
/**
* @brief 主窗口
@ -60,6 +61,7 @@ private:
QMenu* trayMenu_ = nullptr;
QIcon idleIcon_; // SP_MediaStop — 就绪/停止
QIcon activeIcon_; // SP_MediaPlay — 录音/识别
RecordingIndicator* recordingIndicator_ = nullptr;
};
} // namespace impress

12
src/ui/widgets/hotkey_recorder.cpp
View File

@ -1,10 +1,13 @@
#include "hotkey_recorder.h"
#include "utils/logger.h"
#include <QHBoxLayout>
#include <QKeyEvent>
#include <QLabel>
#include <QPushButton>
static const char* const kTag = "HotkeyRecorder";
namespace impress {
HotkeyRecorder::HotkeyRecorder(const QString& label, QWidget* parent)
@ -19,11 +22,12 @@ HotkeyRecorder::HotkeyRecorder(const QString& label, QWidget* parent)
btn_ = new QPushButton(this);
btn_->setMinimumWidth(140);
btn_->setFocusPolicy(Qt::StrongFocus);
btn_->installEventFilter(this); // 拦截按钮的键盘事件
connect(btn_, &QPushButton::clicked, this, &HotkeyRecorder::onToggleRecording);
layout->addWidget(btn_);
layout->addStretch();
setHotkeyText("CapsLock");
setHotkeyText("Ctrl+Alt+C");
applyStyle();
}
@ -44,11 +48,13 @@ void HotkeyRecorder::onToggleRecording() {
if (recording_) {
// 取消录制
recording_ = false;
LOG_INFO(kTag, "录制模式已取消");
updateDisplay();
applyStyle();
} else {
// 进入录制模式
recording_ = true;
LOG_INFO(kTag, "进入录制模式，等待按键...");
updateDisplay();
applyStyle();
btn_->setFocus();
@ -61,15 +67,18 @@ bool HotkeyRecorder::eventFilter(QObject* obj, QEvent* event) {
if (event->type() == QEvent::KeyPress) {
auto* ke = static_cast<QKeyEvent*>(event);
LOG_INFO(kTag, QString("捕获按键: key=%1 modifiers=%2").arg(ke->key()).arg(static_cast<int>(ke->modifiers())));
// 忽略单独的修饰键
if (ke->key() == Qt::Key_Control || ke->key() == Qt::Key_Alt ||
ke->key() == Qt::Key_Shift || ke->key() == Qt::Key_Meta) {
LOG_DEBUG(kTag, "忽略单独修饰键");
return true;
}
// Esc 取消录制
if (ke->key() == Qt::Key_Escape) {
LOG_INFO(kTag, "Esc 取消录制");
recording_ = false;
updateDisplay();
applyStyle();
@ -94,6 +103,7 @@ bool HotkeyRecorder::eventFilter(QObject* obj, QEvent* event) {
hotkeyText_ = text;
recording_ = false;
LOG_INFO(kTag, QString("快捷键已设置: %1").arg(hotkeyText_));
updateDisplay();
applyStyle();
btn_->releaseKeyboard();