Use Opus for VoIP

Server/client VoIP protocol is handled by adding new cvars
cl_voipProtocol and sv_voipProtocol, sv_voip and cl_voip
are used to auto set/clear them. All users need to touch
are cl/sv_voip as 0 or 1 just like before.

Old Speex VoIP packets in demos are skipped.
New VoIP packets are skipped in demos if sv_voipProtocol
doesn't match cl_voipProtocol.

Notable difference between usage of speex and opus codecs,
when using Speex client would be sent 80ms at a time.
Using Opus, 60ms is sent at a time. This was changed because
the Opus codec supports encoding up to 60ms at a time.
(Simpler to send only one codec frame in a packet.)

6 files changed, 127 insertions, 155 deletions

diff --git a/src/client/cl_cgame.c b/src/client/cl_cgame.c
index 1c7d91cc..23f3010c 100644
--- a/src/client/cl_cgame.c
+++ b/src/client/cl_cgame.c
@@ -956,37 +956,27 @@ void CL_FirstSnapshot( void ) {
 #endif
 
 #ifdef USE_VOIP
-	if (!clc.speexInitialized) {
+	if (!clc.voipCodecInitialized) {
 		int i;
-		speex_bits_init(&clc.speexEncoderBits);
-		speex_bits_reset(&clc.speexEncoderBits);
+		int error;
 
-		clc.speexEncoder = speex_encoder_init(&speex_nb_mode);
+		clc.opusEncoder = opus_encoder_create(48000, 1, OPUS_APPLICATION_VOIP, &error);
 
-		speex_encoder_ctl(clc.speexEncoder, SPEEX_GET_FRAME_SIZE,
-		                  &clc.speexFrameSize);
-		speex_encoder_ctl(clc.speexEncoder, SPEEX_GET_SAMPLING_RATE,
-		                  &clc.speexSampleRate);
-
-		clc.speexPreprocessor = speex_preprocess_state_init(clc.speexFrameSize,
-		                                                  clc.speexSampleRate);
-
-		i = 1;
-		speex_preprocess_ctl(clc.speexPreprocessor,
-		                     SPEEX_PREPROCESS_SET_DENOISE, &i);
-
-		i = 1;
-		speex_preprocess_ctl(clc.speexPreprocessor,
-		                     SPEEX_PREPROCESS_SET_AGC, &i);
+		if ( error ) {
+			Com_DPrintf("VoIP: Error opus_encoder_create %d\n", error);
+			return;
+		}
 
 		for (i = 0; i < MAX_CLIENTS; i++) {
-			speex_bits_init(&clc.speexDecoderBits[i]);
-			speex_bits_reset(&clc.speexDecoderBits[i]);
-			clc.speexDecoder[i] = speex_decoder_init(&speex_nb_mode);
+			clc.opusDecoder[i] = opus_decoder_create(48000, 1, &error);
+			if ( error ) {
+				Com_DPrintf("VoIP: Error opus_decoder_create(%d) %d\n", i, error);
+				return;
+			}
 			clc.voipIgnore[i] = qfalse;
 			clc.voipGain[i] = 1.0f;
 		}
-		clc.speexInitialized = qtrue;
+		clc.voipCodecInitialized = qtrue;
 		clc.voipMuteAll = qfalse;
 		Cmd_AddCommand ("voip", CL_Voip_f);
 		Cvar_Set("cl_voipSendTarget", "spatial");
diff --git a/src/client/cl_input.c b/src/client/cl_input.c
index fdd0c474..48bf51df 100644
--- a/src/client/cl_input.c
+++ b/src/client/cl_input.c
@@ -789,7 +789,7 @@ void CL_WritePacket( void ) {
 	{
 		if((clc.voipFlags & VOIP_SPATIAL) || Com_IsVoipTarget(clc.voipTargets, sizeof(clc.voipTargets), -1))
 		{
-			MSG_WriteByte (&buf, clc_voip);
+			MSG_WriteByte (&buf, clc_voipOpus);
 			MSG_WriteByte (&buf, clc.voipOutgoingGeneration);
 			MSG_WriteLong (&buf, clc.voipOutgoingSequence);
 			MSG_WriteByte (&buf, clc.voipOutgoingDataFrames);
@@ -810,7 +810,7 @@ void CL_WritePacket( void ) {
 				MSG_Init (&fakemsg, fakedata, sizeof (fakedata));
 				MSG_Bitstream (&fakemsg);
 				MSG_WriteLong (&fakemsg, clc.reliableAcknowledge);
-				MSG_WriteByte (&fakemsg, svc_voip);
+				MSG_WriteByte (&fakemsg, svc_voipOpus);
 				MSG_WriteShort (&fakemsg, clc.clientNum);
 				MSG_WriteByte (&fakemsg, clc.voipOutgoingGeneration);
 				MSG_WriteLong (&fakemsg, clc.voipOutgoingSequence);
diff --git a/src/client/cl_main.c b/src/client/cl_main.c
index fa47988e..6df57b4d 100644
--- a/src/client/cl_main.c
+++ b/src/client/cl_main.c
@@ -45,6 +45,7 @@ cvar_t	*cl_voipSendTarget;
 cvar_t	*cl_voipGainDuringCapture;
 cvar_t	*cl_voipCaptureMult;
 cvar_t	*cl_voipShowMeter;
+cvar_t	*cl_voipProtocol;
 cvar_t	*cl_voip;
 #endif
 
@@ -250,8 +251,8 @@ void CL_Voip_f( void )
 
 	if (clc.state != CA_ACTIVE)
 		reason = "Not connected to a server";
-	else if (!clc.speexInitialized)
-		reason = "Speex not initialized";
+	else if (!clc.voipCodecInitialized)
+		reason = "Voip codec not initialized";
 	else if (!clc.voipEnabled)
 		reason = "Server doesn't support VoIP";
 
@@ -304,6 +305,8 @@ void CL_VoipNewGeneration(void)
 		clc.voipOutgoingGeneration = 1;
 	clc.voipPower = 0.0f;
 	clc.voipOutgoingSequence = 0;
+
+	opus_encoder_ctl(clc.opusEncoder, OPUS_RESET_STATE);
 }
 
 /*
@@ -392,7 +395,7 @@ void CL_VoipParseTargets(void)
 ===============
 CL_CaptureVoip
 
-Record more audio from the hardware if required and encode it into Speex
+Record more audio from the hardware if required and encode it into Opus
  data for later transmission.
 ===============
 */
@@ -422,11 +425,12 @@ void CL_CaptureVoip(void)
 			Com_Printf("Until then, VoIP is disabled.\n");
 			Cvar_Set("cl_voip", "0");
 		}
+		Cvar_Set("cl_voipProtocol", cl_voip->integer ? "opus" : "");
 		cl_voip->modified = qfalse;
 		cl_rate->modified = qfalse;
 	}
 
-	if (!clc.speexInitialized)
+	if (!clc.voipCodecInitialized)
 		return;  // just in case this gets called at a bad time.
 
 	if (clc.voipOutgoingDataSize > 0)
@@ -479,80 +483,67 @@ void CL_CaptureVoip(void)
 
 	if ((cl_voipSend->integer) || (finalFrame)) { // user wants to capture audio?
 		int samples = S_AvailableCaptureSamples();
-		const int mult = (finalFrame) ? 1 : 4; // 4 == 80ms of audio.
+		const int packetSamples = (finalFrame) ? VOIP_MAX_FRAME_SAMPLES : VOIP_MAX_PACKET_SAMPLES;
 
 		// enough data buffered in audio hardware to process yet?
-		if (samples >= (clc.speexFrameSize * mult)) {
-			// audio capture is always MONO16 (and that's what speex wants!).
-			//  2048 will cover 12 uncompressed frames in narrowband mode.
-			static int16_t sampbuffer[2048];
+		if (samples >= packetSamples) {
+			// audio capture is always MONO16.
+			static int16_t sampbuffer[VOIP_MAX_PACKET_SAMPLES];
 			float voipPower = 0.0f;
-			int speexFrames = 0;
-			int wpos = 0;
-			int pos = 0;
+			int voipFrames;
+			int i, bytes;
 
-			if (samples > (clc.speexFrameSize * 4))
-				samples = (clc.speexFrameSize * 4);
+			if (samples > VOIP_MAX_PACKET_SAMPLES)
+				samples = VOIP_MAX_PACKET_SAMPLES;
 
 			// !!! FIXME: maybe separate recording from encoding, so voipPower
 			// !!! FIXME:  updates faster than 4Hz?
 
-			samples -= samples % clc.speexFrameSize;
-			S_Capture(samples, (byte *) sampbuffer);  // grab from audio card.
-
-			// this will probably generate multiple speex packets each time.
-			while (samples > 0) {
-				int16_t *sampptr = &sampbuffer[pos];
-				int i, bytes;
+			samples -= samples % VOIP_MAX_FRAME_SAMPLES;
+			if (samples != 120 && samples != 240 && samples != 480 && samples != 960 && samples != 1920 && samples != 2880 ) {
+				Com_Printf("Voip: bad number of samples %d\n", samples);
+				return;
+			}
+			voipFrames = samples / VOIP_MAX_FRAME_SAMPLES;
 
-				// preprocess samples to remove noise...
-				speex_preprocess_run(clc.speexPreprocessor, sampptr);
+			S_Capture(samples, (byte *) sampbuffer);  // grab from audio card.
 
-				// check the "power" of this packet...
-				for (i = 0; i < clc.speexFrameSize; i++) {
-					const float flsamp = (float) sampptr[i];
-					const float s = fabs(flsamp);
-					voipPower += s * s;
-					sampptr[i] = (int16_t) ((flsamp) * audioMult);
-				}
+			// check the "power" of this packet...
+			for (i = 0; i < samples; i++) {
+				const float flsamp = (float) sampbuffer[i];
+				const float s = fabs(flsamp);
+				voipPower += s * s;
+				sampbuffer[i] = (int16_t) ((flsamp) * audioMult);
+			}
 
-				// encode raw audio samples into Speex data...
-				speex_bits_reset(&clc.speexEncoderBits);
-				speex_encode_int(clc.speexEncoder, sampptr,
-				                 &clc.speexEncoderBits);
-				bytes = speex_bits_write(&clc.speexEncoderBits,
-				                         (char *) &clc.voipOutgoingData[wpos+1],
-				                         sizeof (clc.voipOutgoingData) - (wpos+1));
-				assert((bytes > 0) && (bytes < 256));
-				clc.voipOutgoingData[wpos] = (byte) bytes;
-				wpos += bytes + 1;
-
-				// look at the data for the next packet...
-				pos += clc.speexFrameSize;
-				samples -= clc.speexFrameSize;
-				speexFrames++;
+			// encode raw audio samples into Opus data...
+			bytes = opus_encode(clc.opusEncoder, sampbuffer, samples,
+									(unsigned char *) clc.voipOutgoingData,
+									sizeof (clc.voipOutgoingData));
+			if ( bytes <= 0 ) {
+				Com_DPrintf("VoIP: Error encoding %d samples\n", samples);
+				bytes = 0;
 			}
 
 			clc.voipPower = (voipPower / (32768.0f * 32768.0f *
-			                 ((float) (clc.speexFrameSize * speexFrames)))) *
-			                 100.0f;
+			                 ((float) samples))) * 100.0f;
 
 			if ((useVad) && (clc.voipPower < cl_voipVADThreshold->value)) {
 				CL_VoipNewGeneration();  // no "talk" for at least 1/4 second.
 			} else {
-				clc.voipOutgoingDataSize = wpos;
-				clc.voipOutgoingDataFrames = speexFrames;
+				clc.voipOutgoingDataSize = bytes;
+				clc.voipOutgoingDataFrames = voipFrames;
 
 				Com_DPrintf("VoIP: Send %d frames, %d bytes, %f power\n",
-				            speexFrames, wpos, clc.voipPower);
+				            voipFrames, bytes, clc.voipPower);
 
 				#if 0
 				static FILE *encio = NULL;
 				if (encio == NULL) encio = fopen("voip-outgoing-encoded.bin", "wb");
-				if (encio != NULL) { fwrite(clc.voipOutgoingData, wpos, 1, encio); fflush(encio); }
+				if (encio != NULL) { fwrite(clc.voipOutgoingData, bytes, 1, encio); fflush(encio); }
 				static FILE *decio = NULL;
 				if (decio == NULL) decio = fopen("voip-outgoing-decoded.bin", "wb");
-				if (decio != NULL) { fwrite(sampbuffer, speexFrames * clc.speexFrameSize * 2, 1, decio); fflush(decio); }
+				if (decio != NULL) { fwrite(sampbuffer, voipFrames * VOIP_MAX_FRAME_SAMPLES * 2, 1, decio); fflush(decio); }
 				#endif
 			}
 		}
@@ -1419,14 +1410,11 @@ void CL_Disconnect( qboolean showMainMenu ) {
 		cl_voipUseVAD->integer = tmp;
 	}
 
-	if (clc.speexInitialized) {
+	if (clc.voipCodecInitialized) {
 		int i;
-		speex_bits_destroy(&clc.speexEncoderBits);
-		speex_encoder_destroy(clc.speexEncoder);
-		speex_preprocess_state_destroy(clc.speexPreprocessor);
+		opus_encoder_destroy(clc.opusEncoder);
 		for (i = 0; i < MAX_CLIENTS; i++) {
-			speex_bits_destroy(&clc.speexDecoderBits[i]);
-			speex_decoder_destroy(clc.speexDecoder[i]);
+			opus_decoder_destroy(clc.opusDecoder[i]);
 		}
 	}
 	Cmd_RemoveCommand ("voip");
@@ -3706,9 +3694,9 @@ void CL_Init( void ) {
 	cl_voipVADThreshold = Cvar_Get ("cl_voipVADThreshold", "0.25", CVAR_ARCHIVE);
 	cl_voipShowMeter = Cvar_Get ("cl_voipShowMeter", "1", CVAR_ARCHIVE);
 
-	// This is a protocol version number.
-	cl_voip = Cvar_Get ("cl_voip", "1", CVAR_USERINFO | CVAR_ARCHIVE);
+	cl_voip = Cvar_Get ("cl_voip", "1", CVAR_ARCHIVE);
 	Cvar_CheckRange( cl_voip, 0, 1, qtrue );
+	cl_voipProtocol = Cvar_Get ("cl_voipProtocol", cl_voip->integer ? "opus" : "", CVAR_USERINFO | CVAR_ROM);
 #endif
 
 
diff --git a/src/client/cl_parse.c b/src/client/cl_parse.c
index a8b68ca5..4230c04f 100644
--- a/src/client/cl_parse.c
+++ b/src/client/cl_parse.c
@@ -35,7 +35,8 @@ char *svc_strings[256] = {
 	"svc_download",
 	"svc_snapshot",
 	"svc_EOF",
-	"svc_voip",
+	"svc_voipSpeex",
+	"svc_voipOpus",
 };
 
 void SHOWNET( msg_t *msg, char *s) {
@@ -354,8 +355,8 @@ void CL_SystemInfoChanged( void ) {
 	cl.serverId = atoi( Info_ValueForKey( systemInfo, "sv_serverid" ) );
 
 #ifdef USE_VOIP
-	s = Info_ValueForKey( systemInfo, "sv_voip" );
-	clc.voipEnabled = atoi(s);
+	s = Info_ValueForKey( systemInfo, "sv_voipProtocol" );
+	clc.voipEnabled = !Q_stricmp(s, "opus");
 #endif
 
 	// don't set any vars when playing a demo
@@ -674,13 +675,13 @@ static void CL_PlayVoip(int sender, int samplecnt, const byte *data, int flags)
 {
 	if(flags & VOIP_DIRECT)
 	{
-		S_RawSamples(sender + 1, samplecnt, clc.speexSampleRate, 2, 1,
+		S_RawSamples(sender + 1, samplecnt, 48000, 2, 1,
 	             data, clc.voipGain[sender], -1);
 	}
 
 	if(flags & VOIP_SPATIAL)
 	{
-		S_RawSamples(sender + MAX_CLIENTS + 1, samplecnt, clc.speexSampleRate, 2, 1,
+		S_RawSamples(sender + MAX_CLIENTS + 1, samplecnt, 48000, 2, 1,
 	             data, 1.0f, sender);
 	}
 }
@@ -693,8 +694,8 @@ A VoIP message has been received from the server
 =====================
 */
 static
-void CL_ParseVoip ( msg_t *msg ) {
-	static short decoded[4096];  // !!! FIXME: don't hardcode.
+void CL_ParseVoip ( msg_t *msg, qboolean ignoreData ) {
+	static short decoded[VOIP_MAX_PACKET_SAMPLES*4]; // !!! FIXME: don't hard code
 
 	const int sender = MSG_ReadShort(msg);
 	const int generation = MSG_ReadByte(msg);
@@ -702,7 +703,8 @@ void CL_ParseVoip ( msg_t *msg ) {
 	const int frames = MSG_ReadByte(msg);
 	const int packetsize = MSG_ReadShort(msg);
 	const int flags = MSG_ReadBits(msg, VOIP_FLAGCNT);
-	char encoded[1024];
+	unsigned char encoded[4000];
+	int	numSamples;
 	int seqdiff;
 	int written = 0;
 	int i;
@@ -732,14 +734,15 @@ void CL_ParseVoip ( msg_t *msg ) {
 		return;   // overlarge packet, bail.
 	}
 
-	if (!clc.speexInitialized) {
-		MSG_ReadData(msg, encoded, packetsize);  // skip payload.
-		return;   // can't handle VoIP without libspeex!
+	MSG_ReadData(msg, encoded, packetsize);
+
+	if (ignoreData) {
+		return; // just ignore legacy speex voip data
+	} else if (!clc.voipCodecInitialized) {
+		return;   // can't handle VoIP without libopus!
 	} else if (sender >= MAX_CLIENTS) {
-		MSG_ReadData(msg, encoded, packetsize);  // skip payload.
 		return;   // bogus sender.
 	} else if (CL_ShouldIgnoreVoipSender(sender)) {
-		MSG_ReadData(msg, encoded, packetsize);  // skip payload.
 		return;   // Channel is muted, bail.
 	}
 
@@ -752,70 +755,59 @@ void CL_ParseVoip ( msg_t *msg ) {
 	// This is a new "generation" ... a new recording started, reset the bits.
 	if (generation != clc.voipIncomingGeneration[sender]) {
 		Com_DPrintf("VoIP: new generation %d!\n", generation);
-		speex_bits_reset(&clc.speexDecoderBits[sender]);
+		opus_decoder_ctl(clc.opusDecoder[sender], OPUS_RESET_STATE);
 		clc.voipIncomingGeneration[sender] = generation;
 		seqdiff = 0;
 	} else if (seqdiff < 0) {   // we're ahead of the sequence?!
 		// This shouldn't happen unless the packet is corrupted or something.
 		Com_DPrintf("VoIP: misordered sequence! %d < %d!\n",
 		            sequence, clc.voipIncomingSequence[sender]);
-		// reset the bits just in case.
-		speex_bits_reset(&clc.speexDecoderBits[sender]);
+		// reset the decoder just in case.
+		opus_decoder_ctl(clc.opusDecoder[sender], OPUS_RESET_STATE);
 		seqdiff = 0;
-	} else if (seqdiff * clc.speexFrameSize * 2 >= sizeof (decoded)) { // dropped more than we can handle?
+	} else if (seqdiff * VOIP_MAX_PACKET_SAMPLES*2 >= sizeof (decoded)) { // dropped more than we can handle?
 		// just start over.
 		Com_DPrintf("VoIP: Dropped way too many (%d) frames from client #%d\n",
 		            seqdiff, sender);
-		speex_bits_reset(&clc.speexDecoderBits[sender]);
+		opus_decoder_ctl(clc.opusDecoder[sender], OPUS_RESET_STATE);
 		seqdiff = 0;
 	}
 
 	if (seqdiff != 0) {
 		Com_DPrintf("VoIP: Dropped %d frames from client #%d\n",
 		            seqdiff, sender);
-		// tell speex that we're missing frames...
+		// tell opus that we're missing frames...
 		for (i = 0; i < seqdiff; i++) {
-			assert((written + clc.speexFrameSize) * 2 < sizeof (decoded));
-			speex_decode_int(clc.speexDecoder[sender], NULL, decoded + written);
-			written += clc.speexFrameSize;
+			assert((written + VOIP_MAX_PACKET_SAMPLES) * 2 < sizeof (decoded));
+			numSamples = opus_decode(clc.opusDecoder[sender], NULL, VOIP_MAX_PACKET_SAMPLES * 2, decoded + written, sizeof (decoded) - written, 0);
+			if ( numSamples <= 0 ) {
+				Com_DPrintf("VoIP: Error decoding frame %d from client #%d\n", i, sender);
+				continue;
+			}
+			written += numSamples;
 		}
 	}
 
-	for (i = 0; i < frames; i++) {
-		const int len = MSG_ReadByte(msg);
-		if (len < 0) {
-			Com_DPrintf("VoIP: Short packet!\n");
-			break;
-		}
-		MSG_ReadData(msg, encoded, len);
+	numSamples = opus_decode(clc.opusDecoder[sender], encoded, packetsize, decoded + written, sizeof (decoded) - written, 0);
 
-		// shouldn't happen, but just in case...
-		if ((written + clc.speexFrameSize) * 2 > sizeof (decoded)) {
-			Com_DPrintf("VoIP: playback %d bytes, %d samples, %d frames\n",
-			            written * 2, written, i);
-			
-			CL_PlayVoip(sender, written, (const byte *) decoded, flags);
-			written = 0;
-		}
-
-		speex_bits_read_from(&clc.speexDecoderBits[sender], encoded, len);
-		speex_decode_int(clc.speexDecoder[sender],
-		                 &clc.speexDecoderBits[sender], decoded + written);
+	if ( numSamples <= 0 ) {
+		Com_DPrintf("VoIP: Error decoding voip data from client #%d\n", sender);
+		numSamples = 0;
+	}
 
-		#if 0
-		static FILE *encio = NULL;
-		if (encio == NULL) encio = fopen("voip-incoming-encoded.bin", "wb");
-		if (encio != NULL) { fwrite(encoded, len, 1, encio); fflush(encio); }
-		static FILE *decio = NULL;
-		if (decio == NULL) decio = fopen("voip-incoming-decoded.bin", "wb");
-		if (decio != NULL) { fwrite(decoded+written, clc.speexFrameSize*2, 1, decio); fflush(decio); }
-		#endif
+	#if 0
+	static FILE *encio = NULL;
+	if (encio == NULL) encio = fopen("voip-incoming-encoded.bin", "wb");
+	if (encio != NULL) { fwrite(encoded, len, 1, encio); fflush(encio); }
+	static FILE *decio = NULL;
+	if (decio == NULL) decio = fopen("voip-incoming-decoded.bin", "wb");
+	if (decio != NULL) { fwrite(decoded+written, clc.speexFrameSize*2, 1, decio); fflush(decio); }
+	#endif
 
-		written += clc.speexFrameSize;
-	}
+	written += numSamples;
 
 	Com_DPrintf("VoIP: playback %d bytes, %d samples, %d frames\n",
-	            written * 2, written, i);
+	            written * 2, written, frames);
 
 	if(written > 0)
 		CL_PlayVoip(sender, written, (const byte *) decoded, flags);
@@ -918,9 +910,14 @@ void CL_ParseServerMessage( msg_t *msg ) {
 		case svc_download:
 			CL_ParseDownload( msg );
 			break;
-		case svc_voip:
+		case svc_voipSpeex:
+#ifdef USE_VOIP
+			CL_ParseVoip( msg, qtrue );
+#endif
+			break;
+		case svc_voipOpus:
 #ifdef USE_VOIP
-			CL_ParseVoip( msg );
+			CL_ParseVoip( msg, !clc.voipEnabled );
 #endif
 			break;
 		}
diff --git a/src/client/client.h b/src/client/client.h
index 541dca40..b4016467 100644
--- a/src/client/client.h
+++ b/src/client/client.h
@@ -36,8 +36,7 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 #endif /* USE_CURL */
 
 #ifdef USE_VOIP
-#include "speex/speex.h"
-#include "speex/speex_preprocess.h"
+#include <opus.h>
 #endif
 
 // file full of random crap that gets used to create cl_guid
@@ -241,14 +240,11 @@ typedef struct {
 
 #ifdef USE_VOIP
 	qboolean voipEnabled;
-	qboolean speexInitialized;
-	int speexFrameSize;
-	int speexSampleRate;
+	qboolean voipCodecInitialized;
 
 	// incoming data...
 	// !!! FIXME: convert from parallel arrays to array of a struct.
-	SpeexBits speexDecoderBits[MAX_CLIENTS];
-	void *speexDecoder[MAX_CLIENTS];
+	OpusDecoder *opusDecoder[MAX_CLIENTS];
 	byte voipIncomingGeneration[MAX_CLIENTS];
 	int voipIncomingSequence[MAX_CLIENTS];
 	float voipGain[MAX_CLIENTS];
@@ -260,9 +256,7 @@ typedef struct {
 	// then we are sending to clientnum i.
 	uint8_t voipTargets[(MAX_CLIENTS + 7) / 8];
 	uint8_t voipFlags;
-	SpeexPreprocessState *speexPreprocessor;
-	SpeexBits speexEncoderBits;
-	void *speexEncoder;
+	OpusEncoder *opusEncoder;
 	int voipOutgoingDataSize;
 	int voipOutgoingDataFrames;
 	int voipOutgoingSequence;
@@ -449,6 +443,13 @@ extern	cvar_t	*cl_voipGainDuringCapture;
 extern	cvar_t	*cl_voipCaptureMult;
 extern	cvar_t	*cl_voipShowMeter;
 extern	cvar_t	*cl_voip;
+
+// 20ms at 48k
+#define VOIP_MAX_FRAME_SAMPLES		( 20 * 48 )
+
+// 3 frame is 60ms of audio, the max opus will encode at once
+#define VOIP_MAX_PACKET_FRAMES		3
+#define VOIP_MAX_PACKET_SAMPLES		( VOIP_MAX_FRAME_SAMPLES * VOIP_MAX_PACKET_FRAMES )
 #endif
 
 //=================================================
diff --git a/src/client/snd_openal.c b/src/client/snd_openal.c
index d90e6d23..b5c89918 100644
--- a/src/client/snd_openal.c
+++ b/src/client/snd_openal.c
@@ -2700,16 +2700,12 @@ qboolean S_AL_Init( soundInterface_t *si )
 
 			s_alAvailableInputDevices = Cvar_Get("s_alAvailableInputDevices", inputdevicenames, CVAR_ROM | CVAR_NORESTART);
 
-			// !!! FIXME: 8000Hz is what Speex narrowband mode needs, but we
-			// !!! FIXME:  should probably open the capture device after
-			// !!! FIXME:  initializing Speex so we can change to wideband
-			// !!! FIXME:  if we like.
 			Com_Printf("OpenAL default capture device is '%s'\n", defaultinputdevice ? defaultinputdevice : "none");
-			alCaptureDevice = qalcCaptureOpenDevice(inputdevice, 8000, AL_FORMAT_MONO16, 4096);
+			alCaptureDevice = qalcCaptureOpenDevice(inputdevice, 48000, AL_FORMAT_MONO16, VOIP_MAX_PACKET_SAMPLES*4);
 			if( !alCaptureDevice && inputdevice )
 			{
 				Com_Printf( "Failed to open OpenAL Input device '%s', trying default.\n", inputdevice );
-				alCaptureDevice = qalcCaptureOpenDevice(NULL, 8000, AL_FORMAT_MONO16, 4096);
+				alCaptureDevice = qalcCaptureOpenDevice(NULL, 48000, AL_FORMAT_MONO16, VOIP_MAX_PACKET_SAMPLES*4);
 			}
 			Com_Printf( "OpenAL capture device %s.\n",
 				    (alCaptureDevice == NULL) ? "failed to open" : "opened");