diff --git a/src/osd/modules/sound/pa_sound.cpp b/src/osd/modules/sound/pa_sound.cpp
index a8f5388ec01..08960cca592 100644
--- a/src/osd/modules/sound/pa_sound.cpp
+++ b/src/osd/modules/sound/pa_sound.cpp
@@ -79,11 +79,12 @@ private:
 	// Used when the stream callbacks need to be protected.
 	// In our case, when audio buffers are manipulated on a running stream.
 	// There is no real need to have one per stream.
-	std::mutex m_stream_mutex; 
+	std::mutex m_stream_mutex;
 
 	std::map<uint32_t, stream_info> m_streams;
 
 	uint32_t m_stream_id;
+	int m_audio_latency;
 
 	int stream_callback(stream_info *stream, const void *input, void *output, unsigned long frameCount, const PaStreamCallbackTimeInfo *timeInfo, PaStreamCallbackFlags statusFlags);
 	static int s_stream_callback(const void *input, void *output, unsigned long frameCount, const PaStreamCallbackTimeInfo *timeInfo, PaStreamCallbackFlags statusFlags, void *userData);
@@ -111,7 +112,7 @@ int sound_pa::init(osd_interface &osd, osd_options const &options)
 		{ -0.2,  0.0,  0.0 },
 		{  0.2,  0.0,  0.0 },
 		{  0.0,  0.0, -1.0 },
-	};		
+	};
 
 	static const uint32_t positions[9][9] = {
 		{ FC },
@@ -136,7 +137,7 @@ int sound_pa::init(osd_interface &osd, osd_options const &options)
 		auto &node = m_info.m_nodes[dev];
 		node.m_name = di->name;
 		node.m_id = dev + 1;
-		node.m_rate.m_default_rate = node.m_rate.m_min_rate = node.m_rate.m_max_rate = di->defaultSampleRate;
+		node.m_rate.m_default_rate = node.m_rate.m_min_rate = node.m_rate.m_max_rate = options.sample_rate();
 		node.m_sinks = di->maxOutputChannels;
 		node.m_sources = di->maxInputChannels;
 
@@ -153,12 +154,13 @@ int sound_pa::init(osd_interface &osd, osd_options const &options)
 	m_info.m_default_sink = dc(Pa_GetDefaultOutputDevice());
 	m_info.m_default_source = dc(Pa_GetDefaultInputDevice());
 
+	m_audio_latency = options.audio_latency();
 	m_stream_id = 1;
 
 	return 0;
 }
 
-void sound_pa::exit() 
+void sound_pa::exit()
 {
 	Pa_Terminate();
 }
@@ -188,7 +190,7 @@ uint32_t sound_pa::stream_sink_open(uint32_t node, std::string name, uint32_t ra
 	op.device = node - 1;
 	op.channelCount = m_info.m_nodes[node-1].m_sinks;
 	op.sampleFormat = paInt16;
-	op.suggestedLatency = Pa_GetDeviceInfo(node - 1)->defaultLowOutputLatency;
+	op.suggestedLatency = (m_audio_latency > 0) ? (m_audio_latency / 10.0) : Pa_GetDeviceInfo(node - 1)->defaultLowOutputLatency;
 	op.hostApiSpecificStreamInfo = nullptr;
 
 	PaError err = Pa_OpenStream(&si->second.m_stream, nullptr, &op, rate, paFramesPerBufferUnspecified, 0, s_stream_callback, &si->second);
@@ -216,7 +218,7 @@ uint32_t sound_pa::stream_source_open(uint32_t node, std::string name, uint32_t
 	ip.device = node - 1;
 	ip.channelCount = m_info.m_nodes[node-1].m_sources;
 	ip.sampleFormat = paInt16;
-	ip.suggestedLatency = Pa_GetDeviceInfo(node - 1)->defaultLowInputLatency;
+	ip.suggestedLatency = (m_audio_latency > 0) ? (m_audio_latency / 10.0) : Pa_GetDeviceInfo(node - 1)->defaultLowInputLatency;
 	ip.hostApiSpecificStreamInfo = nullptr;
 
 	PaError err = Pa_OpenStream(&si->second.m_stream, &ip, nullptr, rate, paFramesPerBufferUnspecified, 0, s_stream_callback, &si->second);
@@ -224,7 +226,7 @@ uint32_t sound_pa::stream_source_open(uint32_t node, std::string name, uint32_t
 		err = Pa_SetStreamFinishedCallback(si->second.m_stream, s_stream_finished_callback);
 	if(!err)
 		err = Pa_StartStream(si->second.m_stream);
-	
+
 	if(err) {
 		m_streams.erase(si);
 		return 0;