diff --git a/src/single_window/render.cpp b/src/single_window/render.cpp
index 311a382..dd50c88 100644
--- a/src/single_window/render.cpp
+++ b/src/single_window/render.cpp
@@ -370,8 +370,9 @@ int Render::StopScreenCapturer() {
 int Render::StartSpeakerCapturer() {
   if (!speaker_capturer_) {
     speaker_capturer_ = (SpeakerCapturer*)speaker_capturer_factory_->Create();
-    int speaker_capturer_init_ret = speaker_capturer_->Init(
-        [this](unsigned char* data, size_t size) -> void {
+    int speaker_capturer_init_ret =
+        speaker_capturer_->Init([this](unsigned char* data, size_t size,
+                                       const char* audio_name) -> void {
           SendAudioFrame(peer_, (const char*)data, size, audio_label_.c_str());
         });
 
diff --git a/src/speaker_capturer/macosx/speaker_capturer_macosx.cpp b/src/speaker_capturer/macosx/speaker_capturer_macosx.cpp
index c33f74c..e69de29 100644
--- a/src/speaker_capturer/macosx/speaker_capturer_macosx.cpp
+++ b/src/speaker_capturer/macosx/speaker_capturer_macosx.cpp
@@ -1,14 +0,0 @@
-
-#include "speaker_capturer_macosx.h"
-
-SpeakerCapturerMacosx::SpeakerCapturerMacosx() {}
-
-SpeakerCapturerMacosx::~SpeakerCapturerMacosx() {}
-
-int SpeakerCapturerMacosx::Init(speaker_data_cb cb) { return 0; }
-int SpeakerCapturerMacosx::Destroy() { return 0; }
-int SpeakerCapturerMacosx::Start() { return 0; }
-int SpeakerCapturerMacosx::Stop() { return 0; }
-
-int SpeakerCapturerMacosx::Pause() { return 0; }
-int SpeakerCapturerMacosx::Resume() { return 0; }
\ No newline at end of file
diff --git a/src/speaker_capturer/macosx/speaker_capturer_macosx.h b/src/speaker_capturer/macosx/speaker_capturer_macosx.h
index 2ffe2aa..54d23df 100644
--- a/src/speaker_capturer/macosx/speaker_capturer_macosx.h
+++ b/src/speaker_capturer/macosx/speaker_capturer_macosx.h
@@ -26,13 +26,12 @@ class SpeakerCapturerMacosx : public SpeakerCapturer {
   int Pause();
   int Resume();
 
- private:
+ public:
   speaker_data_cb cb_ = nullptr;
-
- private:
   bool inited_ = false;
-  // thread
-  std::thread capture_thread_;
+
+  class Impl;
+  Impl* impl_ = nullptr;
 };
 
 #endif
\ No newline at end of file
diff --git a/src/speaker_capturer/macosx/speaker_capturer_macosx.mm b/src/speaker_capturer/macosx/speaker_capturer_macosx.mm
new file mode 100644
index 0000000..1d91065
--- /dev/null
+++ b/src/speaker_capturer/macosx/speaker_capturer_macosx.mm
@@ -0,0 +1,210 @@
+#import <AVFoundation/AVFoundation.h>
+#import <Foundation/Foundation.h>
+#import <ScreenCaptureKit/ScreenCaptureKit.h>
+
+#include "speaker_capturer_macosx.h"
+
+// 这个delegate用来接收SCStream回调
+@interface SpeakerCaptureDelegate : NSObject <SCStreamDelegate, SCStreamOutput>
+@property(nonatomic, assign) SpeakerCapturerMacosx* owner;  // assign用于C++指针，不用weak
+- (instancetype)initWithOwner:(SpeakerCapturerMacosx*)owner;
+@end
+
+@implementation SpeakerCaptureDelegate
+- (instancetype)initWithOwner:(SpeakerCapturerMacosx*)owner {
+  self = [super init];
+  if (self) {
+    _owner = owner;
+  }
+  return self;
+}
+
+- (void)stream:(SCStream*)stream
+    didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer
+                   ofType:(SCStreamOutputType)type {
+  if (type == SCStreamOutputTypeAudio) {
+    CMBlockBufferRef blockBuffer = CMSampleBufferGetDataBuffer(sampleBuffer);
+    size_t length = CMBlockBufferGetDataLength(blockBuffer);
+    char* dataPtr = NULL;
+    CMBlockBufferGetDataPointer(blockBuffer, 0, NULL, NULL, &dataPtr);
+
+    // 获取输入格式
+    CMAudioFormatDescriptionRef formatDesc = CMSampleBufferGetFormatDescription(sampleBuffer);
+    const AudioStreamBasicDescription* asbd =
+        CMAudioFormatDescriptionGetStreamBasicDescription(formatDesc);
+
+    if (_owner->cb_ && dataPtr && length > 0 && asbd) {
+      std::vector<short> out_pcm16;
+      if (asbd->mFormatFlags & kAudioFormatFlagIsFloat) {
+        int channels = asbd->mChannelsPerFrame;
+        int samples = (int)(length / sizeof(float));
+        float* floatData = (float*)dataPtr;
+        std::vector<short> pcm16(samples);
+        for (int i = 0; i < samples; ++i) {
+          float v = floatData[i];
+          if (v > 1.0f) v = 1.0f;
+          if (v < -1.0f) v = -1.0f;
+          pcm16[i] = (short)(v * 32767.0f);
+        }
+        // 混合为单声道
+        if (channels > 1) {
+          int mono_samples = samples / channels;
+          out_pcm16.resize(mono_samples);
+          for (int i = 0; i < mono_samples; ++i) {
+            int sum = 0;
+            for (int c = 0; c < channels; ++c) {
+              sum += pcm16[i * channels + c];
+            }
+            out_pcm16[i] = sum / channels;
+          }
+        } else {
+          out_pcm16 = std::move(pcm16);
+        }
+      } else if (asbd->mBitsPerChannel == 16) {
+        int channels = asbd->mChannelsPerFrame;
+        int samples = (int)(length / 2);
+        short* src = (short*)dataPtr;
+        if (channels > 1) {
+          int mono_samples = samples / channels;
+          out_pcm16.resize(mono_samples);
+          for (int i = 0; i < mono_samples; ++i) {
+            int sum = 0;
+            for (int c = 0; c < channels; ++c) {
+              sum += src[i * channels + c];
+            }
+            out_pcm16[i] = sum / channels;
+          }
+        } else {
+          out_pcm16.assign(src, src + samples);
+        }
+      }
+
+      // 分包，每960字节送一次cb_（即480采样点）
+      size_t frame_bytes = 960;  // 480 * 2
+      size_t total_bytes = out_pcm16.size() * sizeof(short);
+      unsigned char* p = (unsigned char*)out_pcm16.data();
+      for (size_t offset = 0; offset + frame_bytes <= total_bytes; offset += frame_bytes) {
+        _owner->cb_(p + offset, frame_bytes, "audio");
+      }
+      // 如有剩余，可缓存到下次补齐
+    }
+  }
+}
+@end
+
+// C++类实现细节，放这里，隐藏在.mm中
+class SpeakerCapturerMacosx::Impl {
+ public:
+  SCStreamConfiguration* config = nil;
+  SCStream* stream = nil;
+  SpeakerCaptureDelegate* delegate = nil;
+};
+
+SpeakerCapturerMacosx::SpeakerCapturerMacosx() { impl_ = new Impl(); }
+
+SpeakerCapturerMacosx::~SpeakerCapturerMacosx() {
+  Destroy();
+  delete impl_;
+  impl_ = nullptr;
+}
+
+int SpeakerCapturerMacosx::Init(speaker_data_cb cb) {
+  if (inited_) return 0;
+  cb_ = cb;
+
+  impl_->config = [[SCStreamConfiguration alloc] init];
+  impl_->config.capturesAudio = YES;
+  impl_->config.sampleRate = 48000;
+  impl_->config.channelCount = 1;
+
+  impl_->delegate = [[SpeakerCaptureDelegate alloc] initWithOwner:this];
+
+  // 异步获取可共享内容，改为同步等待
+  dispatch_semaphore_t sema = dispatch_semaphore_create(0);
+  __block SCShareableContent* content = nil;
+  __block NSError* error = nil;
+  [SCShareableContent
+      getShareableContentWithCompletionHandler:^(SCShareableContent* c, NSError* e) {
+        content = c;
+        error = e;
+        dispatch_semaphore_signal(sema);
+      }];
+  dispatch_semaphore_wait(sema, DISPATCH_TIME_FOREVER);
+
+  if (error || !content) return -1;
+
+  // 选择主显示屏
+  CGDirectDisplayID mainDisplayId = CGMainDisplayID();
+  SCDisplay* mainDisplay = nil;
+  for (SCDisplay* d in content.displays) {
+    if (d.displayID == mainDisplayId) {
+      mainDisplay = d;
+      break;
+    }
+  }
+  if (!mainDisplay) return -1;
+
+  // 用SCContentFilter包装
+  SCContentFilter* filter = [[SCContentFilter alloc] initWithDisplay:mainDisplay
+                                                    excludingWindows:@[]];
+
+  impl_->stream = [[SCStream alloc] initWithFilter:filter
+                                     configuration:impl_->config
+                                          delegate:impl_->delegate];
+
+  NSError* addOutputError = nil;
+  dispatch_queue_t queue = dispatch_queue_create("SpeakerAudio.Queue", DISPATCH_QUEUE_SERIAL);
+  BOOL ok = [impl_->stream addStreamOutput:impl_->delegate
+                                      type:SCStreamOutputTypeAudio
+                        sampleHandlerQueue:queue
+                                     error:&addOutputError];
+  if (!ok || addOutputError) {
+    NSLog(@"addStreamOutput error: %@", addOutputError);
+    return -1;
+  }
+
+  inited_ = true;
+  return 0;
+}
+
+int SpeakerCapturerMacosx::Start() {
+  if (!inited_) return -1;
+  dispatch_semaphore_t sema = dispatch_semaphore_create(0);
+  __block int ret = 0;
+  [impl_->stream startCaptureWithCompletionHandler:^(NSError* _Nullable error) {
+    if (error) ret = -1;
+    dispatch_semaphore_signal(sema);
+  }];
+  dispatch_semaphore_wait(sema, DISPATCH_TIME_FOREVER);
+  return ret;
+}
+
+int SpeakerCapturerMacosx::Stop() {
+  if (!inited_) return -1;
+  dispatch_semaphore_t sema = dispatch_semaphore_create(0);
+  [impl_->stream stopCaptureWithCompletionHandler:^(NSError* error) {
+    dispatch_semaphore_signal(sema);
+  }];
+  dispatch_semaphore_wait(sema, DISPATCH_TIME_FOREVER);
+  inited_ = false;
+  return 0;
+}
+
+int SpeakerCapturerMacosx::Destroy() {
+  Stop();
+  cb_ = nullptr;
+  if (impl_) {
+    impl_->delegate = nil;
+    impl_->stream = nil;
+    impl_->config = nil;
+  }
+  inited_ = false;
+  return 0;
+}
+
+int SpeakerCapturerMacosx::Pause() {
+  // ScreenCaptureKit无暂停接口，暂时无实现
+  return 0;
+}
+
+int SpeakerCapturerMacosx::Resume() { return Start(); }
diff --git a/src/speaker_capturer/speaker_capturer.h b/src/speaker_capturer/speaker_capturer.h
index 3166409..a53bcfe 100644
--- a/src/speaker_capturer/speaker_capturer.h
+++ b/src/speaker_capturer/speaker_capturer.h
@@ -11,7 +11,8 @@
 
 class SpeakerCapturer {
  public:
-  typedef std::function<void(unsigned char *, size_t)> speaker_data_cb;
+  typedef std::function<void(unsigned char *, size_t, const char *)>
+      speaker_data_cb;
 
  public:
   virtual ~SpeakerCapturer() {}
diff --git a/src/speaker_capturer/windows/speaker_capturer_wasapi.cpp b/src/speaker_capturer/windows/speaker_capturer_wasapi.cpp
index 46ca733..8a94c6c 100644
--- a/src/speaker_capturer/windows/speaker_capturer_wasapi.cpp
+++ b/src/speaker_capturer/windows/speaker_capturer_wasapi.cpp
@@ -24,7 +24,8 @@ void data_callback(ma_device* pDevice, void* pOutput, const void* pInput,
     }
 
     ptr->GetCallback()((unsigned char*)pInput,
-                       frameCount * ma_get_bytes_per_frame(format_, channels_));
+                       frameCount * ma_get_bytes_per_frame(format_, channels_),
+                       "audio");
   }
 
   (void)pOutput;
diff --git a/thirdparty/minirtc b/thirdparty/minirtc
index b44495e..9b8134d 160000
--- a/thirdparty/minirtc
+++ b/thirdparty/minirtc
@@ -1 +1 @@
-Subproject commit b44495ef609198ff14a3e0e171c0ba4f664da8b8
+Subproject commit 9b8134d48757c07df7ee0e9dbc6d5dc6f36fb0b2
diff --git a/xmake.lua b/xmake.lua
index 7a9c358..471ca8e 100644
--- a/xmake.lua
+++ b/xmake.lua
@@ -37,7 +37,7 @@ elseif is_os("macosx") then
     add_ldflags("-Wl,-ld_classic")
     add_cxflags("-Wno-unused-variable")
     add_frameworks("OpenGL", "IOSurface", "ScreenCaptureKit", "AVFoundation", 
-        "CoreMedia", "CoreVideo")
+        "CoreMedia", "CoreVideo", "CoreAudio", "AudioToolbox")
 end
 
 add_packages("spdlog", "imgui")
@@ -83,7 +83,8 @@ target("speaker_capturer")
         add_files("src/speaker_capturer/windows/*.cpp")
         add_includedirs("src/speaker_capturer/windows", {public = true})
     elseif is_os("macosx") then
-        add_files("src/speaker_capturer/macosx/*.cpp")
+        add_files("src/speaker_capturer/macosx/*.cpp",
+        "src/speaker_capturer/macosx/*.mm")
         add_includedirs("src/speaker_capturer/macosx", {public = true})
     elseif is_os("linux") then
         add_files("src/speaker_capturer/linux/*.cpp")