diff --git a/audio/out/ao_audiounit.m b/audio/out/ao_audiounit.m
index 70a80f29f6..677b2281d2 100644
--- a/audio/out/ao_audiounit.m
+++ b/audio/out/ao_audiounit.m
@@ -49,8 +49,8 @@ static OSStatus render_cb_lpcm(void *ctx, AudioUnitRenderActionFlags *aflags,
         planes[n] = buffer_list->mBuffers[n].mData;
 
     int64_t end = mp_time_us();
-    end += ca_frames_to_us(ao, frames);
     end += p->device_latency * 1e6;
+    end += ca_get_latency(ts) + ca_frames_to_us(ao, frames);
     ao_read_data(ao, planes, frames, end);
     return noErr;
 }
diff --git a/audio/out/ao_coreaudio_utils.c b/audio/out/ao_coreaudio_utils.c
index d458b581ae..b55bf1c0df 100644
--- a/audio/out/ao_coreaudio_utils.c
+++ b/audio/out/ao_coreaudio_utils.c
@@ -31,6 +31,8 @@
 #if HAVE_COREAUDIO
 #include "audio/out/ao_coreaudio_properties.h"
 #include <CoreAudio/HostTime.h>
+#else
+#include <mach/mach_time.h>
 #endif
 
 CFStringRef cfstr_from_cstr(char *str)
@@ -310,9 +312,9 @@ int64_t ca_frames_to_us(struct ao *ao, uint32_t frames)
     return frames / (float) ao->samplerate * 1e6;
 }
 
-#if HAVE_COREAUDIO
 int64_t ca_get_latency(const AudioTimeStamp *ts)
 {
+#if HAVE_COREAUDIO
     uint64_t out = AudioConvertHostTimeToNanos(ts->mHostTime);
     uint64_t now = AudioConvertHostTimeToNanos(AudioGetCurrentHostTime());
 
@@ -320,8 +322,22 @@ int64_t ca_get_latency(const AudioTimeStamp *ts)
         return 0;
 
     return (out - now) * 1e-3;
+#else
+    static mach_timebase_info_data_t timebase;
+    if (timebase.denom == 0)
+        mach_timebase_info(&timebase);
+
+    uint64_t out = ts->mHostTime;
+    uint64_t now = mach_absolute_time();
+
+    if (now > out)
+        return 0;
+
+    return (out - now) * timebase.numer / timebase.denom / 1e3;
+#endif
 }
 
+#if HAVE_COREAUDIO
 bool ca_stream_supports_compressed(struct ao *ao, AudioStreamID stream)
 {
     AudioStreamRangedDescription *formats = NULL;
diff --git a/audio/out/ao_coreaudio_utils.h b/audio/out/ao_coreaudio_utils.h
index 68f5b076ec..cd1f5f282f 100644
--- a/audio/out/ao_coreaudio_utils.h
+++ b/audio/out/ao_coreaudio_utils.h
@@ -65,9 +65,7 @@ bool ca_asbd_is_better(AudioStreamBasicDescription *req,
                        AudioStreamBasicDescription *new);
 
 int64_t ca_frames_to_us(struct ao *ao, uint32_t frames);
-#if HAVE_COREAUDIO
 int64_t ca_get_latency(const AudioTimeStamp *ts);
-#endif
 
 #if HAVE_COREAUDIO
 bool ca_stream_supports_compressed(struct ao *ao, AudioStreamID stream);