diff --git a/daemon/codec.c b/daemon/codec.c
index a1b5f0fa3..24a312456 100644
--- a/daemon/codec.c
+++ b/daemon/codec.c
@@ -1066,7 +1066,7 @@ static int packet_encoded_t38(encoder_t *enc, void *u1, void *u2) {
 		return 0;
 
 	return t38_gateway_input_samples(mp->media->t38_gateway,
-			(int16_t *) enc->avpkt.data, enc->avpkt.size / 2);
+			(int16_t *) enc->avpkt->data, enc->avpkt->size / 2);
 }
 
 static void __generator_stop(struct call_media *media) {
@@ -2913,14 +2913,14 @@ static int packet_encoded_rtp(encoder_t *enc, void *u1, void *u2) {
 	//unsigned int seq_off = (mp->iter_out > mp->iter_in) ? 1 : 0;
 
 	ilogs(transcoding, LOG_DEBUG, "RTP media successfully encoded: TS %llu, len %i",
-			(unsigned long long) enc->avpkt.pts, enc->avpkt.size);
+			(unsigned long long) enc->avpkt->pts, enc->avpkt->size);
 
 	// run this through our packetizer
-	AVPacket *in_pkt = &enc->avpkt;
+	AVPacket *in_pkt = enc->avpkt;
 
 	while (1) {
 		// figure out how big of a buffer we need
-		unsigned int payload_len = MAX(MAX(enc->avpkt.size, ch->bytes_per_packet),
+		unsigned int payload_len = MAX(MAX(enc->avpkt->size, ch->bytes_per_packet),
 				sizeof(struct telephone_event_payload));
 		unsigned int pkt_len = sizeof(struct rtp_header) + payload_len + RTP_BUFFER_TAIL_ROOM;
 		// prepare our buffers
@@ -2935,7 +2935,7 @@ static int packet_encoded_rtp(encoder_t *enc, void *u1, void *u2) {
 		int ret = enc->def->packetizer(in_pkt,
 				ch->sample_buffer, &inout, enc);
 
-		if (G_UNLIKELY(ret == -1 || enc->avpkt.pts == AV_NOPTS_VALUE)) {
+		if (G_UNLIKELY(ret == -1 || enc->avpkt->pts == AV_NOPTS_VALUE)) {
 			// nothing
 			free(buf);
 			break;
@@ -2948,7 +2948,7 @@ static int packet_encoded_rtp(encoder_t *enc, void *u1, void *u2) {
 		unsigned int repeats = 0;
 		int payload_type = -1;
 
-		int is_dtmf = dtmf_event_payload(&inout, (uint64_t *) &enc->avpkt.pts, enc->avpkt.duration,
+		int is_dtmf = dtmf_event_payload(&inout, (uint64_t *) &enc->avpkt->pts, enc->avpkt->duration,
 				&ch->dtmf_event, &ch->dtmf_events);
 		if (is_dtmf) {
 			payload_type = ch->handler->dtmf_payload_type;
@@ -2958,7 +2958,7 @@ static int packet_encoded_rtp(encoder_t *enc, void *u1, void *u2) {
 				repeats = 2; // DTMF end event
 		}
 		else {
-			if (is_silence_event(&inout, &ch->silence_events, enc->avpkt.pts, enc->avpkt.duration))
+			if (is_silence_event(&inout, &ch->silence_events, enc->avpkt->pts, enc->avpkt->duration))
 				payload_type = ch->handler->cn_payload_type;
 		}
 
@@ -2972,7 +2972,7 @@ static int packet_encoded_rtp(encoder_t *enc, void *u1, void *u2) {
 				memcpy(send_buf, buf, pkt_len);
 			}
 			__output_rtp(mp, ch, ch->handler, send_buf, inout.len, ch->first_ts
-					+ enc->avpkt.pts / enc->def->clockrate_mult,
+					+ enc->avpkt->pts / enc->def->clockrate_mult,
 					ch->rtp_mark ? 1 : 0, -1, 0,
 					payload_type, 0);
 			mp->ssrc_out->parent->seq_diff++;
diff --git a/daemon/media_player.c b/daemon/media_player.c
index 1d29ac816..30aae353e 100644
--- a/daemon/media_player.c
+++ b/daemon/media_player.c
@@ -98,6 +98,7 @@ static void __media_player_free(void *p) {
 	ssrc_ctx_put(&mp->ssrc_out);
 	mutex_destroy(&mp->lock);
 	obj_put(mp->call);
+	av_packet_free(&mp->pkt);
 }
 #endif
 
@@ -124,9 +125,9 @@ struct media_player *media_player_new(struct call_monologue *ml) {
 	mp->seq = ssl_random();
 	mp->ssrc_out = ssrc_ctx;
 
-	av_init_packet(&mp->pkt);
-	mp->pkt.data = NULL;
-	mp->pkt.size = 0;
+	mp->pkt = av_packet_alloc();
+	mp->pkt->data = NULL;
+	mp->pkt->size = 0;
 
 	return mp;
 #else
@@ -388,7 +389,7 @@ static void media_player_read_packet(struct media_player *mp) {
 	if (!mp->fmtctx)
 		return;
 
-	int ret = av_read_frame(mp->fmtctx, &mp->pkt);
+	int ret = av_read_frame(mp->fmtctx, mp->pkt);
 	if (ret < 0) {
 		if (ret == AVERROR_EOF) {
 			if (mp->repeat > 1){
@@ -400,7 +401,7 @@ static void media_player_read_packet(struct media_player *mp) {
 				ret = av_seek_frame(mp->fmtctx, -1, 0, 0);
 				if (ret < 0)
 					ilog(LOG_ERR, "Failed to seek to beginning of media file");
-				ret = av_read_frame(mp->fmtctx, &mp->pkt);
+				ret = av_read_frame(mp->fmtctx, mp->pkt);
 			} else {
 				ilog(LOG_DEBUG, "EOF reading from media stream");
 				return;
@@ -431,26 +432,26 @@ static void media_player_read_packet(struct media_player *mp) {
 
 	// scale pts and duration according to sample rate
 
-	long long duration_scaled = mp->pkt.duration * avs->CODECPAR->sample_rate
+	long long duration_scaled = mp->pkt->duration * avs->CODECPAR->sample_rate
 		* avs->time_base.num / avs->time_base.den;
-	unsigned long long pts_scaled = mp->pkt.pts * avs->CODECPAR->sample_rate
+	unsigned long long pts_scaled = mp->pkt->pts * avs->CODECPAR->sample_rate
 		* avs->time_base.num / avs->time_base.den;
 
-	long long us_dur = mp->pkt.duration * 1000000LL * avs->time_base.num / avs->time_base.den;
+	long long us_dur = mp->pkt->duration * 1000000LL * avs->time_base.num / avs->time_base.den;
 	ilog(LOG_DEBUG, "read media packet: pts %llu duration %lli (scaled %llu/%lli, %lli us), "
 			"sample rate %i, time_base %i/%i",
-			(unsigned long long) mp->pkt.pts,
-			(long long) mp->pkt.duration,
+			(unsigned long long) mp->pkt->pts,
+			(long long) mp->pkt->duration,
 			pts_scaled,
 			duration_scaled,
 			us_dur,
 			avs->CODECPAR->sample_rate,
 			avs->time_base.num, avs->time_base.den);
 
-	media_player_add_packet(mp, (char *) mp->pkt.data, mp->pkt.size, us_dur, pts_scaled);
+	media_player_add_packet(mp, (char *) mp->pkt->data, mp->pkt->size, us_dur, pts_scaled);
 
 out:
-	av_packet_unref(&mp->pkt);
+	av_packet_unref(mp->pkt);
 }
 
 
diff --git a/include/media_player.h b/include/media_player.h
index bf8f5693a..53bdeca73 100644
--- a/include/media_player.h
+++ b/include/media_player.h
@@ -43,7 +43,7 @@ struct media_player {
 
 	AVFormatContext *fmtctx;
 	unsigned long duration; // in milliseconds
-	AVPacket pkt;
+	AVPacket *pkt;
 	struct codec_handler *handler;
 	struct ssrc_ctx *ssrc_out;
 	unsigned long seq;
diff --git a/lib/codeclib.c b/lib/codeclib.c
index f13f5f4a1..afc29f743 100644
--- a/lib/codeclib.c
+++ b/lib/codeclib.c
@@ -590,6 +590,8 @@ static const char *avc_decoder_init(decoder_t *dec, const str *fmtp, const str *
 	if (!codec)
 		return "codec not supported";
 
+	dec->u.avc.avpkt = av_packet_alloc();
+
 	dec->u.avc.avcctx = avcodec_alloc_context3(codec);
 	if (!dec->u.avc.avcctx)
 		return "failed to alloc codec context";
@@ -663,8 +665,6 @@ decoder_t *decoder_new_fmtp(const codec_def_t *def, int clockrate, int channels,
 	if (err)
 		goto err;
 
-	av_init_packet(&ret->u.avc.avpkt);
-
 	ret->pts = (uint64_t) -1LL;
 	ret->rtp_ts = (unsigned long) -1L;
 
@@ -722,6 +722,7 @@ static void avc_decoder_close(decoder_t *dec) {
 	avcodec_close(dec->u.avc.avcctx);
 	av_free(dec->u.avc.avcctx);
 #endif
+	av_packet_free(&dec->u.avc.avpkt);
 }
 
 
@@ -744,9 +745,9 @@ static int avc_decoder_input(decoder_t *dec, const str *data, GQueue *out) {
 	const char *err;
 	int av_ret = 0;
 
-	dec->u.avc.avpkt.data = (unsigned char *) data->s;
-	dec->u.avc.avpkt.size = data->len;
-	dec->u.avc.avpkt.pts = dec->pts;
+	dec->u.avc.avpkt->data = (unsigned char *) data->s;
+	dec->u.avc.avpkt->size = data->len;
+	dec->u.avc.avpkt->pts = dec->pts;
 
 	AVFrame *frame = NULL;
 
@@ -761,13 +762,13 @@ static int avc_decoder_input(decoder_t *dec, const str *data, GQueue *out) {
 			goto err;
 
 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(57, 36, 0)
-		if (dec->u.avc.avpkt.size) {
-			av_ret = avcodec_send_packet(dec->u.avc.avcctx, &dec->u.avc.avpkt);
+		if (dec->u.avc.avpkt->size) {
+			av_ret = avcodec_send_packet(dec->u.avc.avcctx, dec->u.avc.avpkt);
 			cdbg("send packet ret %i", av_ret);
 			err = "failed to send packet to avcodec";
 			if (av_ret == 0) {
 				// consumed the packet
-				dec->u.avc.avpkt.size = 0;
+				dec->u.avc.avpkt->size = 0;
 				keep_going = 1;
 			}
 			else {
@@ -794,10 +795,10 @@ static int avc_decoder_input(decoder_t *dec, const str *data, GQueue *out) {
 		}
 #else
 		// only do this if we have any input left
-		if (dec->u.avc.avpkt.size == 0)
+		if (dec->u.avc.avpkt->size == 0)
 			break;
 
-		av_ret = avcodec_decode_audio4(dec->u.avc.avcctx, frame, &got_frame, &dec->u.avc.avpkt);
+		av_ret = avcodec_decode_audio4(dec->u.avc.avcctx, frame, &got_frame, dec->u.avc.avpkt);
 		cdbg("decode frame ret %i, got frame %i", av_ret, got_frame);
 		err = "failed to decode audio packet";
 		if (av_ret < 0)
@@ -805,10 +806,10 @@ static int avc_decoder_input(decoder_t *dec, const str *data, GQueue *out) {
 		if (av_ret > 0) {
 			// consumed some input
 			err = "invalid return value";
-			if (av_ret > dec->u.avc.avpkt.size)
+			if (av_ret > dec->u.avc.avpkt->size)
 				goto err;
-			dec->u.avc.avpkt.size -= av_ret;
-			dec->u.avc.avpkt.data += av_ret;
+			dec->u.avc.avpkt->size -= av_ret;
+			dec->u.avc.avpkt->data += av_ret;
 			keep_going = 1;
 		}
 		if (got_frame)
@@ -823,8 +824,8 @@ static int avc_decoder_input(decoder_t *dec, const str *data, GQueue *out) {
 			frame->pts = frame->pkt_pts;
 #endif
 			if (G_UNLIKELY(frame->pts == AV_NOPTS_VALUE))
-				frame->pts = dec->u.avc.avpkt.pts;
-			dec->u.avc.avpkt.pts += frame->nb_samples;
+				frame->pts = dec->u.avc.avpkt->pts;
+			dec->u.avc.avpkt->pts += frame->nb_samples;
 
 			g_queue_push_tail(out, frame);
 			frame = NULL;
@@ -1311,7 +1312,7 @@ int encoder_config_fmtp(encoder_t *enc, const codec_def_t *def, int bitrate, int
 	if (err)
 		goto err;
 
-	av_init_packet(&enc->avpkt);
+	enc->avpkt = av_packet_alloc();
 
 // output frame and fifo
 	enc->frame = av_frame_alloc();
@@ -1370,6 +1371,7 @@ void encoder_close(encoder_t *enc) {
 }
 void encoder_free(encoder_t *enc) {
 	encoder_close(enc);
+	av_packet_free(&enc->avpkt);
 	g_slice_free1(sizeof(*enc), enc);
 }
 
@@ -1398,7 +1400,7 @@ static int avc_encoder_input(encoder_t *enc, AVFrame **frame) {
 		}
 	}
 
-	av_ret = avcodec_receive_packet(enc->u.avc.avcctx, &enc->avpkt);
+	av_ret = avcodec_receive_packet(enc->u.avc.avcctx, enc->avpkt);
 	cdbg("receive packet ret %i", av_ret);
 	if (av_ret == 0) {
 		// got some data
@@ -1415,7 +1417,7 @@ static int avc_encoder_input(encoder_t *enc, AVFrame **frame) {
 	if (!*frame)
 		return 0;
 
-	av_ret = avcodec_encode_audio2(enc->u.avc.avcctx, &enc->avpkt, *frame, &got_packet);
+	av_ret = avcodec_encode_audio2(enc->u.avc.avcctx, enc->avpkt, *frame, &got_packet);
 	cdbg("encode frame ret %i, got packet %i", av_ret, got_packet);
 	if (av_ret == 0)
 		*frame = NULL; // consumed
@@ -1428,16 +1430,16 @@ static int avc_encoder_input(encoder_t *enc, AVFrame **frame) {
 	if (!got_packet)
 		return keep_going;
 
-	cdbg("output avpkt size is %i", (int) enc->avpkt.size);
-	cdbg("output pkt pts/dts is %li/%li", (long) enc->avpkt.pts,
-			(long) enc->avpkt.dts);
+	cdbg("output avpkt size is %i", (int) enc->avpkt->size);
+	cdbg("output pkt pts/dts is %li/%li", (long) enc->avpkt->pts,
+			(long) enc->avpkt->dts);
 
 	// the encoder may return frames with the same dts multiple consecutive times.
 	// the muxer may not like this, so ensure monotonically increasing dts.
-	if (enc->mux_dts > enc->avpkt.dts)
-		enc->avpkt.dts = enc->mux_dts;
-	if (enc->avpkt.pts < enc->avpkt.dts)
-		enc->avpkt.pts = enc->avpkt.dts;
+	if (enc->mux_dts > enc->avpkt->dts)
+		enc->avpkt->dts = enc->mux_dts;
+	if (enc->avpkt->pts < enc->avpkt->dts)
+		enc->avpkt->pts = enc->avpkt->dts;
 
 	return keep_going;
 
@@ -1450,7 +1452,7 @@ err:
 int encoder_input_data(encoder_t *enc, AVFrame *frame,
 		int (*callback)(encoder_t *, void *u1, void *u2), void *u1, void *u2)
 {
-	enc->avpkt.size = 0;
+	enc->avpkt->size = 0;
 
 	while (1) {
 		if (!enc->def->codec_type->encoder_input)
@@ -1460,22 +1462,22 @@ int encoder_input_data(encoder_t *enc, AVFrame *frame,
 		if (ret < 0)
 			return -1;
 
-		if (enc->avpkt.size) {
+		if (enc->avpkt->size) {
 			// don't rely on the encoder producing steady timestamps,
 			// instead keep track of them ourselves based on the returned
 			// frame duration
-			enc->avpkt.pts = enc->next_pts;
+			enc->avpkt->pts = enc->next_pts;
 
 			if (enc->def->codec_type->encoder_got_packet)
 				enc->def->codec_type->encoder_got_packet(enc);
 
 			callback(enc, u1, u2);
 
-			enc->next_pts += enc->avpkt.duration;
-			enc->mux_dts = enc->avpkt.dts + 1; // min next expected dts
+			enc->next_pts += enc->avpkt->duration;
+			enc->mux_dts = enc->avpkt->dts + 1; // min next expected dts
 
-			av_packet_unref(&enc->avpkt);
-			enc->avpkt.size = 0;
+			av_packet_unref(enc->avpkt);
+			enc->avpkt->size = 0;
 		}
 
 		if (ret == 0)
@@ -1550,7 +1552,7 @@ static int packetizer_samplestream(AVPacket *pkt, GString *buf, str *input_outpu
 	memcpy(input_output->s, buf->str, input_output->len);
 	g_string_erase(buf, 0, input_output->len);
 	// adjust output pts
-	enc->avpkt.pts = enc->packet_pts;
+	enc->avpkt->pts = enc->packet_pts;
 	enc->packet_pts += input_output->len * (enc->def->bits_per_sample * enc->def->clockrate_mult / 8);
 	return buf->len >= input_output->len ? 1 : 0;
 }
@@ -2339,8 +2341,8 @@ static int generic_silence_dtx(decoder_t *dec, GQueue *out, int ptime) {
 	memset(frame->extended_data[0], 0, frame->linesize[0]);
 
 	// advance PTS
-	frame->pts = dec->u.avc.avpkt.pts;
-	dec->u.avc.avpkt.pts += frame->nb_samples;
+	frame->pts = dec->u.avc.avpkt->pts;
+	dec->u.avc.avpkt->pts += frame->nb_samples;
 
 	g_queue_push_tail(out, frame);
 
@@ -2454,17 +2456,17 @@ static int bcg729_encoder_input(encoder_t *enc, AVFrame **frame) {
 		return -1;
 	}
 
-	av_new_packet(&enc->avpkt, 10);
+	av_new_packet(enc->avpkt, 10);
 	unsigned char len = 0;
 
-	bcg729Encoder(enc->u.bcg729, (void *) (*frame)->extended_data[0], enc->avpkt.data, &len);
+	bcg729Encoder(enc->u.bcg729, (void *) (*frame)->extended_data[0], enc->avpkt->data, &len);
 	if (!len) {
-		av_packet_unref(&enc->avpkt);
+		av_packet_unref(enc->avpkt);
 		return 0;
 	}
 
-	enc->avpkt.size = len;
-	enc->avpkt.pts = (*frame)->pts;
+	enc->avpkt->size = len;
+	enc->avpkt->pts = (*frame)->pts;
 
 	return 0;
 }
diff --git a/lib/codeclib.h b/lib/codeclib.h
index ed698e367..9ea719d22 100644
--- a/lib/codeclib.h
+++ b/lib/codeclib.h
@@ -212,7 +212,7 @@ struct decoder_s {
 	union {
 		struct {
 			AVCodecContext *avcctx;
-			AVPacket avpkt;
+			AVPacket *avpkt;
 
 			union {
 				struct {
@@ -265,7 +265,7 @@ struct encoder_s {
 		bcg729EncoderChannelContextStruct *bcg729;
 #endif
 	} u;
-	AVPacket avpkt;
+	AVPacket *avpkt;
 	AVAudioFifo *fifo;
 	int64_t fifo_pts; // pts of first data in fifo
 	int64_t packet_pts; // first pts of data in packetizer buffer
diff --git a/recording-daemon/output.c b/recording-daemon/output.c
index 2797098ca..166c09e89 100644
--- a/recording-daemon/output.c
+++ b/recording-daemon/output.c
@@ -27,12 +27,12 @@ static int output_shutdown(output_t *output);
 static int output_got_packet(encoder_t *enc, void *u1, void *u2) {
 	output_t *output = u1;
 
-	dbg("{%s%s%s} output avpkt size is %i", FMT_M(output->file_name), (int) enc->avpkt.size);
-	dbg("{%s%s%s} output pkt pts/dts is %li/%li", FMT_M(output->file_name), (long) enc->avpkt.pts,
-			(long) enc->avpkt.dts);
+	dbg("{%s%s%s} output avpkt size is %i", FMT_M(output->file_name), (int) enc->avpkt->size);
+	dbg("{%s%s%s} output pkt pts/dts is %li/%li", FMT_M(output->file_name), (long) enc->avpkt->pts,
+			(long) enc->avpkt->dts);
 	dbg("{%s%s%s} output dts %li", FMT_M(output->file_name), (long) output->encoder->mux_dts);
 
-	av_write_frame(output->fmtctx, &enc->avpkt);
+	av_write_frame(output->fmtctx, enc->avpkt);
 
 	return 0;
 }