diff --git a/src/audio_stream.c b/src/audio_stream.c
index 542129e..ce4be5d 100644
--- a/src/audio_stream.c
+++ b/src/audio_stream.c
@@ -39,6 +39,7 @@ static long speaker_freq_ctr = 0;
 static u8 audio_volume = AUDIO_VOLUME_MAX;
 static double audio_prev_time;
 static int audio_freq;
+static bool audio_signed;
 
 #ifdef AUDIO_STREAM_DEBUG
 static void audio_stream_print(int pos) {
@@ -71,10 +72,11 @@ static void audio_stream_flag_speaker_overrun(void) {
 	}
 }
 
-void audio_stream_init(long time, int freq) {
+void audio_stream_init(long time, int freq, bool asigned) {
 	speaker_overrun_flagged = 0;
 	audio_prev_time = -1;
 	audio_freq = freq;
+	audio_signed = asigned;
 }
 
 u8 audio_stream_get_volume() {
@@ -102,10 +104,14 @@ void audio_stream_generate_u8(long time, u8 *stream, int len) {
 	double audio_dfrom, audio_dto;
 	long audio_from, audio_to, audio_last_to = 0;
 
+	u8 audio_smp_min = audio_signed ? (-audio_volume) : (128 - audio_volume);
+	u8 audio_smp_max = audio_signed ? audio_volume : (128 + audio_volume);
+	u8 audio_smp_center = audio_signed ? 0 : 128;
+
 	// handle the first
 	if (audio_prev_time < 0) {
 		audio_prev_time = time;
-		memset(stream, 128, len);
+		memset(stream, audio_smp_center, len);
 		speaker_entry_pos = 0;
 		return;
 	}
@@ -125,7 +131,7 @@ void audio_stream_generate_u8(long time, u8 *stream, int len) {
 
 	if (speaker_entry_pos == 0) {
 		audio_curr_time = time;
-		memset(stream, 128, len);
+		memset(stream, audio_smp_center, len);
 	} else {
 		double last_ms = audio_prev_time;
 		for (i = 0; i < speaker_entry_pos; i++) {
@@ -173,9 +179,9 @@ void audio_stream_generate_u8(long time, u8 *stream, int len) {
 					pos_samples_fixed = (speaker_freq_ctr << 8) % (freq_samples_fixed << 1);
 					for (j = audio_from; j < audio_to; j++) {
 						if ((pos_samples_fixed & 0xFFFFFF00) < (freq_samples_fixed & 0xFFFFFF00))
-							stream[j] = 128+audio_volume;
+							stream[j] = audio_smp_max;
 						else
-							stream[j] = 128-audio_volume;
+							stream[j] = audio_smp_min;
 						pos_samples_fixed = (pos_samples_fixed + 256) % (freq_samples_fixed << 1);
 					}
 					speaker_freq_ctr += audio_to - audio_from;
@@ -184,7 +190,7 @@ void audio_stream_generate_u8(long time, u8 *stream, int len) {
 					fprintf(stderr, "[callback] emitting off (%ld, %ld)\n", audio_from, audio_to);
 					#endif
 					speaker_freq_ctr = 0;
-					memset(stream + audio_from, 128, audio_to - audio_from);
+					memset(stream + audio_from, audio_smp_center, audio_to - audio_from);
 				}
 			}
 
diff --git a/src/audio_stream.h b/src/audio_stream.h
index efc81c0..e163a98 100644
--- a/src/audio_stream.h
+++ b/src/audio_stream.h
@@ -26,7 +26,7 @@
 #include "types.h"
 
 USER_FUNCTION
-void audio_stream_init(long time, int freq);
+void audio_stream_init(long time, int freq, bool asigned);
 USER_FUNCTION
 u8 audio_stream_get_volume();
 USER_FUNCTION
diff --git a/src/psp/frontend.c b/src/psp/frontend.c
index 4200bee..1852fcd 100644
--- a/src/psp/frontend.c
+++ b/src/psp/frontend.c
@@ -276,12 +276,11 @@ void psp_audio_callback(void *stream, unsigned int len, void *userdata) {
 		return;
 	}
 
-	u8 *stream_u8 = ((u8*) stream) + (len * 3);
+	s8 *stream_s8 = ((s8*) stream) + (len * 3);
 	s16 *stream_s16 = ((s16*) stream);
 
-	audio_stream_generate_u8(zeta_time_ms(), stream_u8, len);
-	for (int i = 0; i < len; i++, stream_u8++, stream_s16+=2) {
-		s8 sample_s8 = (s8) (stream_u8[0] ^ 0x80);
+	audio_stream_generate_u8(zeta_time_ms(), (u8*) stream_s8, len);
+	for (int i = 0; i < len; i++, stream_s8++, stream_s16+=2) {
 		s16 val = ((s16) sample_s8) << 8;
 		stream_s16[0] = (s16) val;
 		stream_s16[1] = (s16) val;
diff --git a/src/sdl/frontend.c b/src/sdl/frontend.c
index 4549ed0..b95d0a5 100644
--- a/src/sdl/frontend.c
+++ b/src/sdl/frontend.c
@@ -367,7 +367,7 @@ int main(int argc, char **argv) {
 	}
 
 	if (audio_device != 0) {
-		audio_stream_init(zeta_time_ms(), audio_spec.freq);
+		audio_stream_init(zeta_time_ms(), audio_spec.freq, audio_spec.format == AUDIO_S8);
 		if (posix_zzt_arg_note_delay >= 0.0) {
 			audio_set_note_delay(posix_zzt_arg_note_delay);
 		}
diff --git a/src/types.h b/src/types.h
index 2e9a652..101f2f5 100644
--- a/src/types.h
+++ b/src/types.h
@@ -23,6 +23,8 @@
 #ifndef __TYPES_H__
 #define __TYPES_H__
 
+#include <stdbool.h>
+
 #ifdef __EMSCRIPTEN__
 #define USER_FUNCTION EMSCRIPTEN_KEEPALIVE
 #define IMPLEMENT_FUNCTION
diff --git a/web/src/audio.js b/web/src/audio.js
index ba5e7ac..28cca7e 100644
--- a/web/src/audio.js
+++ b/web/src/audio.js
@@ -156,7 +156,7 @@ export class BufferBasedAudio {
 
 		this.time = audioCtx.currentTime;
 
-		this.emu._audio_stream_init(time_ms(), this.sampleRate);
+		this.emu._audio_stream_init(time_ms(), this.sampleRate, false);
 		this.emu._audio_stream_set_volume(Math.floor(this.volume * this.emu._audio_stream_get_max_volume()));
 
 		if (this.noteDelay) {