diff --git a/daemon/audio_player.c b/daemon/audio_player.c
index a5fba28b5..326ab743a 100644
--- a/daemon/audio_player.c
+++ b/daemon/audio_player.c
@@ -112,7 +112,7 @@ bool audio_player_setup(struct call_media *m, const rtp_payload_type *dst_pt,
 	if (mp)
 		media_player_stop(mp);
 	else {
-		media_player_new(&mp, m->monologue, NULL, NULL);
+		media_player_new(&mp, m, NULL, NULL);
 		ap->mp = mp;
 	}
 	if (!mp)
@@ -170,7 +170,7 @@ void audio_player_start(struct call_media *m) {
 	if (!mp)
 		return;
 
-	media_player_set_media(mp, m);
+	media_player_set_sink(mp);
 
 	if (mp->next_run) // already running?
 		return;
diff --git a/daemon/call.c b/daemon/call.c
index 5538a53fd..e3e84acad 100644
--- a/daemon/call.c
+++ b/daemon/call.c
@@ -68,7 +68,6 @@ unsigned int call_socket_cpu_affinity = 0;
 static int64_t add_ongoing_calls_dur_in_interval(int64_t interval_start, int64_t interval_duration);
 static void __call_free(call_t *p);
 static void __call_cleanup(call_t *c);
-static void __monologue_stop(struct call_monologue *ml);
 __attribute__((nonnull(1, 2, 4)))
 static struct media_subscription *__subscribe_medias_both_ways(struct call_media * a, struct call_media * b,
 		bool is_offer, medias_q *);
@@ -2716,8 +2715,7 @@ static void __update_init_subscribers(struct call_media *media, struct stream_pa
 		mqtt_timer_start(&media->mqtt_timer, media->call, media);
 }
 
-__attribute__((nonnull(1)))
-static void update_init_subscribers(struct call_media *media, struct stream_params *sp,
+void update_init_subscribers(struct call_media *media, struct stream_params *sp,
 		sdp_ng_flags *flags, enum ng_opmode opmode)
 {
 	__update_init_subscribers(media, sp, flags, opmode, ++media->call->update_iter);
@@ -5186,6 +5184,8 @@ static void __call_cleanup(call_t *c) {
 		ice_shutdown(&md->ice_agent);
 		call_media_stop(md);
 		t38_gateway_put(&md->t38_gateway);
+		media_player_put(&md->players[MP_DEFAULT]);
+		media_player_put(&md->players[MP_REC]);
 		audio_player_free(md);
 		mutex_destroy(&md->dtmf_lock);
 		sdp_sp_clear(&md->sp);
@@ -5193,9 +5193,6 @@ static void __call_cleanup(call_t *c) {
 
 	for (__auto_type l = c->monologues.head; l; l = l->next) {
 		struct call_monologue *ml = l->data;
-		__monologue_stop(ml);
-		media_player_put(&ml->players[MP_DEFAULT]);
-		media_player_put(&ml->players[MP_REC]);
 		if (ml->tone_freqs)
 			g_array_free(ml->tone_freqs, true);
 		obj_release(ml->janus_session);
@@ -6428,25 +6425,21 @@ void call_media_stop(struct call_media *m) {
 	if (!m)
 		return;
 	t38_gateway_stop(m->t38_gateway);
+	media_player_stop(m->players[MP_DEFAULT]);
+	media_player_stop(m->players[MP_REC]);
 	audio_player_stop(m);
 	codec_handlers_stop(&m->codec_handlers_store, NULL, false);
 	rtcp_timer_stop(&m->rtcp_timer);
 	mqtt_timer_stop(&m->mqtt_timer);
 }
-/**
- * Stops media player of given monologue.
- */
-static void __monologue_stop(struct call_monologue *ml) {
-	media_player_stop(ml->players[MP_DEFAULT]);
-	media_player_stop(ml->players[MP_REC]);
-}
+
+
 /**
  * Stops media player and all medias of given monologue.
  * If asked, stops all media subscribers as well.
  */
 static void monologue_stop(struct call_monologue *ml, bool stop_media_subscribers) {
 	/* monologue itself */
-	__monologue_stop(ml);
 	for (unsigned int i = 0; i < ml->medias->len; i++)
 	{
 		call_media_stop(ml->medias->pdata[i]);
@@ -6458,10 +6451,8 @@ static void monologue_stop(struct call_monologue *ml, bool stop_media_subscriber
 			struct call_media *media = ml->medias->pdata[i];
 			if (!media)
 				continue;
-			IQUEUE_FOREACH(&media->media_subscribers, ms) {
+			IQUEUE_FOREACH(&media->media_subscribers, ms)
 				call_media_stop(ms->media);
-				__monologue_stop(ms->monologue);
-			}
 		}
 	}
 }
diff --git a/daemon/call_interfaces.c b/daemon/call_interfaces.c
index 510519b10..4ff6bfb45 100644
--- a/daemon/call_interfaces.c
+++ b/daemon/call_interfaces.c
@@ -2007,8 +2007,9 @@ const char *call_play_media_ng(ng_command_ctx_t *ctx) {
 		if (err)
 			return err;
 
-		if (l == monologues.head && monologue->players[MP_DEFAULT]->coder.duration)
-			parser->dict_add_int(ctx->resp, "duration", monologue->players[MP_DEFAULT]->coder.duration);
+		if (l == monologues.head && monologue->audio
+				&& monologue->audio->players[MP_DEFAULT]->coder.duration)
+			parser->dict_add_int(ctx->resp, "duration", monologue->audio->players[MP_DEFAULT]->coder.duration);
 
 	}
 
@@ -2034,11 +2035,15 @@ const char *call_stop_media_ng(ng_command_ctx_t *ctx) {
 
 	for (__auto_type l = monologues.head; l; l = l->next) {
 		struct call_monologue *monologue = l->data;
+		struct call_media *audio = monologue->audio;
 
-		if (!monologue->players[MP_DEFAULT])
+		if (!audio)
 			return "Not currently playing media";
 
-		if (monologue->players[MP_DEFAULT]->opts.moh)
+		if (!audio->players[MP_DEFAULT])
+			return "Not currently playing media";
+
+		if (audio->players[MP_DEFAULT]->opts.moh)
 			return "Currently MoH ongoing, ignore stop media.";
 
 		last_frame_pos = call_stop_media_for_ml(monologue, MP_DEFAULT);
diff --git a/daemon/codec.c b/daemon/codec.c
index 733788ff3..9a376cbf5 100644
--- a/daemon/codec.c
+++ b/daemon/codec.c
@@ -1507,7 +1507,7 @@ void __codec_handlers_update(struct call_media *source, struct call_media *sink,
 
 	if (rtpe_config.use_audio_player == UAP_PLAY_MEDIA) {
 		// check for implicitly enabled player
-		if ((a.flags && a.flags->opmode == OP_PLAY_MEDIA) || (media_player_is_active(sink_ml))) {
+		if ((a.flags && a.flags->opmode == OP_PLAY_MEDIA) || (media_player_is_active(sink))) {
 			use_audio_player = true;
 			implicit_audio_player = true;
 		}
diff --git a/daemon/media_player.c b/daemon/media_player.c
index 48aaf3775..5169c4192 100644
--- a/daemon/media_player.c
+++ b/daemon/media_player.c
@@ -183,7 +183,6 @@ static void media_player_shutdown(struct media_player *mp) {
 	if (mp->opts.block_egress && mp->media)
 		MEDIA_CLEAR(mp->media, BLOCK_EGRESS);
 
-	mp->media = NULL;
 	media_player_coder_shutdown(&mp->coder);
 
 	if (mp->kernel_idx != KERNEL_IDX_NONE)
@@ -232,7 +231,7 @@ static void __media_player_free(struct media_player *mp) {
 
 
 // call->master_lock held in W
-void media_player_new(struct media_player **mpp, struct call_monologue *ml, struct ssrc_entry_call *prev_ssrc,
+void media_player_new(struct media_player **mpp, struct call_media *media, struct ssrc_entry_call *prev_ssrc,
 		media_player_opts_t *opts)
 {
 #ifdef WITH_TRANSCODING
@@ -251,8 +250,8 @@ void media_player_new(struct media_player **mpp, struct call_monologue *ml, stru
 		mp->kernel_idx = KERNEL_IDX_NONE;
 
 		mp->run_func = media_player_read_packet; // default
-		mp->call = obj_get(ml->call);
-		mp->ml = ml;
+		mp->call = obj_get(media->call);
+		mp->media = media;
 		if (prev_ssrc)
 			mp->seq = atomic_get_na(&prev_ssrc->stats->ext_seq) + 1;
 		else
@@ -1060,7 +1059,7 @@ static int __ensure_codec_handler(struct media_player *mp, const rtp_payload_typ
 	src_pt.channels = GET_CHANNELS(mp->coder.avstream->codecpar);
 	src_pt.clock_rate = mp->coder.avstream->codecpar->sample_rate;
 
-	codec_init_payload_type(&src_pt, MT_AUDIO);
+	codec_init_payload_type(&src_pt, mp->media->type_id);
 
 	if (__media_player_setup_internal(mp, &src_pt, dst_pt, codec_set))
 		return -1;
@@ -1205,8 +1204,8 @@ static bool media_player_read_packet(struct media_player *mp) {
 
 
 // call->master_lock held in W
-void media_player_set_media(struct media_player *mp, struct call_media *media) {
-	mp->media = media;
+void media_player_set_sink(struct media_player *mp) {
+	struct call_media *media = mp->media;
 	if (media->streams.head) {
 		mp->sink.sink = media->streams.head->data;
 		sink_handler_set_generic(&mp->sink);
@@ -1226,26 +1225,7 @@ void media_player_set_media(struct media_player *mp, struct call_media *media) {
 // call->master_lock held in W
 // returns destination payload type, or NULL on failure
 static const rtp_payload_type *media_player_play_setup(struct media_player *mp) {
-	// find call media suitable for playback
-	struct call_media *media;
-	for (unsigned int i = 0; i < mp->ml->medias->len; i++) {
-		media = mp->ml->medias->pdata[i];
-		if (media->type_id != MT_AUDIO)
-			continue;
-		/* moh allows inactive */
-		if (!mp->opts.moh && !MEDIA_ISSET(media, SEND))
-			continue;
-		if (media->streams.length == 0)
-			continue;
-		goto found;
-	}
-	media = NULL;
-found:
-	if (!media) {
-		ilog(LOG_ERR, "No suitable SDP section for media playback");
-		return NULL;
-	}
-	media_player_set_media(mp, media);
+	media_player_set_sink(mp);
 	return media_player_get_dst_pt(mp);
 }
 
@@ -1581,11 +1561,18 @@ static bool call_ml_stops_moh(struct call_monologue *from_ml, struct call_monolo
 		enum ng_opmode opmode)
 {
 #ifdef WITH_TRANSCODING
-	if (opmode == OP_OFFER && !call_ml_sendonly_inactive(from_ml)
-			&& (to_ml->players[MP_DEFAULT] && to_ml->players[MP_DEFAULT]->opts.moh))
-	{
-		return true;
-	}
+	if (opmode != OP_OFFER)
+		return false;
+	if (call_ml_sendonly_inactive(from_ml))
+		return false;
+	if (!to_ml->audio)
+		return false;
+	if (!to_ml->audio->players[MP_DEFAULT])
+		return false;
+	if (!to_ml->audio->players[MP_DEFAULT]->opts.moh)
+		return false;
+
+	return true;
 #endif
 	return false;
 }
@@ -1601,48 +1588,48 @@ static bool call_ml_stops_moh(struct call_monologue *from_ml, struct call_monolo
  */
 static void call_ml_moh_handle_flags(struct call_monologue *from_ml, struct call_monologue *to_ml) {
 #ifdef WITH_TRANSCODING
-
 	/* if from_ml not given, then it's a reflected MoH, use capabilities of to_ml */
 	struct call_monologue *moh_ml = from_ml ? : to_ml;
 
-	if (!to_ml->players[MP_DEFAULT] ||
+	struct call_media *audio = to_ml->audio;
+	if (!audio)
+		return;
+
+	if (!audio->players[MP_DEFAULT] ||
 		!ML_ISSET2(moh_ml, MOH_ZEROCONN, MOH_SENDRECV))
 	{
 		return;
 	}
 
-	struct call_media * media = to_ml->players[MP_DEFAULT]->media;
-	if (media) {
-		/* check zero-connection */
-		if (ML_ISSET(moh_ml, MOH_ZEROCONN)) {
-			struct packet_stream *ps;
-			__auto_type msl = media->streams.head;
-			while (msl)
-			{
-				ps = msl->data;
-				if (PS_ISSET(ps, RTP)) { /* find RTP stream, and don't touch RTCP */
-					ilog(LOG_DEBUG, "Forced packet stream of '"STR_FORMAT"' (media index: '%d')"
-							"to zero_addr due to MoH zero-connection.",
-							STR_FMT(&media->monologue->tag), media->index);
-					PS_SET(ps, ZERO_ADDR);
-					goto check_next; /* stop */
-				}
-				msl = msl->next;
+	/* check zero-connection */
+	if (ML_ISSET(moh_ml, MOH_ZEROCONN)) {
+		struct packet_stream *ps;
+		__auto_type msl = audio->streams.head;
+		while (msl)
+		{
+			ps = msl->data;
+			if (PS_ISSET(ps, RTP)) { /* find RTP stream, and don't touch RTCP */
+				ilog(LOG_DEBUG, "Forced packet stream of '"STR_FORMAT"' (media index: '%d')"
+						"to zero_addr due to MoH zero-connection.",
+						STR_FMT(&audio->monologue->tag), audio->index);
+				PS_SET(ps, ZERO_ADDR);
+				goto check_next; /* stop */
 			}
+			msl = msl->next;
 		}
+	}
 check_next:
-		/* check mode sendrecv */
-		if (ML_ISSET(moh_ml, MOH_SENDRECV)) {
-			bf_set(&media->media_flags, MEDIA_FLAG_SEND | MEDIA_FLAG_RECV);
-			bf_set(&media->media_flags, MEDIA_FLAG_FAKE_SENDRECV);
-
-			if (media->media_subscriptions.head) {
-				__auto_type sub = media->media_subscriptions.head;
-				__auto_type sub_m = sub->media;
-				/* mark real state of originators media (no flag set - inactive, real_sendonly - sendonly) */
-				if (MEDIA_ISSET(sub_m, RECV))
-					bf_set(&media->media_flags, MEDIA_FLAG_REAL_SENDONLY);
-			}
+	/* check mode sendrecv */
+	if (ML_ISSET(moh_ml, MOH_SENDRECV)) {
+		bf_set(&audio->media_flags, MEDIA_FLAG_SEND | MEDIA_FLAG_RECV);
+		bf_set(&audio->media_flags, MEDIA_FLAG_FAKE_SENDRECV);
+
+		if (audio->media_subscriptions.head) {
+			__auto_type sub = audio->media_subscriptions.head;
+			__auto_type sub_m = sub->media;
+			/* mark real state of originators media (no flag set - inactive, real_sendonly - sendonly) */
+			if (MEDIA_ISSET(sub_m, RECV))
+				bf_set(&audio->media_flags, MEDIA_FLAG_REAL_SENDONLY);
 		}
 	}
 #endif
@@ -1681,10 +1668,8 @@ const char *call_check_moh(struct call_monologue *from_ml, struct call_monologue
 
 		/* whom to play the moh audio */
 		errstr = call_play_media_for_ml(to_ml, MP_DEFAULT, &opts, NULL);
-		if (errstr) {
-			to_ml->players[MP_DEFAULT]->opts.moh = 0; /* initialization failed, mark accordingly */
+		if (errstr)
 			return errstr;
-		}
 
 		/* handle MoH related flags */
 		call_ml_moh_handle_flags((reflected ? NULL : from_ml), to_ml);
@@ -1703,6 +1688,27 @@ const char *call_check_moh(struct call_monologue *from_ml, struct call_monologue
 #endif
 }
 
+
+static void set_monologue_media(struct call_monologue *ml, bool allow_inactive) {
+	ml->audio = NULL;
+
+	for (unsigned int i = 0; i < ml->medias->len; i++) {
+		__auto_type m = ml->medias->pdata[i];
+		if (!m)
+			continue;
+
+		if (m->streams.length == 0)
+			continue;
+
+		if (!ml->audio && m->type_id == MT_AUDIO)
+			if (allow_inactive || MEDIA_ISSET(m, SEND))
+				ml->audio = m;
+
+		if (ml->audio)
+			return;
+	}
+}
+
 const char *call_play_media_for_ml(struct call_monologue *ml, unsigned int mp_idx,
 		media_player_opts_t *opts, sdp_ng_flags *flags)
 {
@@ -1713,7 +1719,13 @@ const char *call_play_media_for_ml(struct call_monologue *ml, unsigned int mp_id
 	/* this starts the audio player if needed */
 	update_init_monologue_subscribers(ml, OP_PLAY_MEDIA);
 
-	if (ml->players[mp_idx] && ml->players[mp_idx]->opts.moh) {
+	set_monologue_media(ml, opts->moh);
+
+	struct call_media *audio = ml->audio;
+	if (!audio)
+		return "No suitable output for media playback";
+
+	if (audio->players[mp_idx] && audio->players[mp_idx]->opts.moh) {
 		ilog(LOG_DEBUG, "There is already ongoing media playback for MoH. Ignore new one.");
 		/* pretend that everything is good */
 		return NULL;
@@ -1721,11 +1733,13 @@ const char *call_play_media_for_ml(struct call_monologue *ml, unsigned int mp_id
 	else {
 		/* media_player_new() now knows that audio player is in use
 		* TODO: player options can have changed if already exists */
-		media_player_new(&ml->players[mp_idx], ml, NULL, opts);
+		media_player_new(&audio->players[mp_idx], audio, NULL, opts);
 	}
 
-	if (!media_player_play(ml->players[mp_idx], opts))
+	if (!media_player_play(audio->players[mp_idx], opts)) {
+		audio->players[mp_idx]->opts.moh = 0;
 		return "Failed to start media playback";
+	}
 
 	return NULL;
 #else
@@ -1733,17 +1747,29 @@ const char *call_play_media_for_ml(struct call_monologue *ml, unsigned int mp_id
 #endif
 }
 
-long long call_stop_media_for_ml(struct call_monologue *ml, unsigned int mp_idx)
-{
+
 #ifdef WITH_TRANSCODING
-	if (!ml->players[mp_idx])
+static long long call_stop_media(struct call_media *m, unsigned int mp_idx) {
+	if (!m)
+		return 0;
+	if (!m->players[mp_idx])
 		return 0;
-	long long ret = media_player_stop(ml->players[mp_idx]);
+
+	long long ret = media_player_stop(m->players[mp_idx]);
 	/* restore to non-mixing if needed */
-	codec_update_all_source_handlers(ml);
-	update_init_monologue_subscribers(ml, OP_STOP_MEDIA);
+	codec_update_media_source_handlers(m);
+	update_init_subscribers(m, NULL, NULL, OP_STOP_MEDIA);
 	/* mark MoH as already not used (it can be unset now) */
-	ml->players[mp_idx]->opts.moh = 0;
+	m->players[mp_idx]->opts.moh = 0;
+	return ret;
+}
+#endif
+
+
+long long call_stop_media_for_ml(struct call_monologue *ml, unsigned int mp_idx)
+{
+#ifdef WITH_TRANSCODING
+	long long ret = call_stop_media(ml->audio, mp_idx);
 	return ret;
 #else
 	return 0;
@@ -2052,12 +2078,12 @@ static void media_player_run(void *ptr) {
 }
 
 
-bool media_player_is_active(struct call_monologue *ml) {
-	if (!ml)
+bool media_player_is_active(struct call_media *m) {
+	if (!m)
 		return false;
-	if (!ml->players[MP_DEFAULT])
+	if (!m->players[MP_DEFAULT])
 		return false;
-	if (!ml->players[MP_DEFAULT]->next_run)
+	if (!m->players[MP_DEFAULT]->next_run)
 		return false;
 	return true;
 }
diff --git a/daemon/mqtt.c b/daemon/mqtt.c
index 835c196cf..e58ff30a2 100644
--- a/daemon/mqtt.c
+++ b/daemon/mqtt.c
@@ -169,35 +169,6 @@ static void mqtt_monologue_stats(struct call_monologue *ml, JsonBuilder *json) {
 		json_builder_set_member_name(json, "label");
 		glib_json_builder_add_str(json, &ml->label);
 	}
-
-#ifdef WITH_TRANSCODING
-	struct media_player *mp = ml->players[MP_DEFAULT];
-	if (mp) {
-		mutex_lock(&mp->lock);
-
-		json_builder_set_member_name(json, "media_player");
-
-		json_builder_begin_object(json);
-
-		json_builder_set_member_name(json, "duration");
-		json_builder_add_int_value(json, mp->coder.duration);
-		json_builder_set_member_name(json, "repeat");
-		json_builder_add_int_value(json, mp->opts.repeat);
-		json_builder_set_member_name(json, "frame_time");
-		json_builder_add_int_value(json, mp->last_frame_ts);
-
-		if (mp->ssrc_out && mp->media) {
-			json_builder_set_member_name(json, "SSRC");
-			json_builder_begin_object(json);
-			mqtt_ssrc_stats(mp->ssrc_out, json, mp->media);
-			json_builder_end_object(json);
-		}
-
-		json_builder_end_object(json);
-
-		mutex_unlock(&mp->lock);
-	}
-#endif
 }
 
 
@@ -465,6 +436,35 @@ static void mqtt_media_stats(struct call_media *media, JsonBuilder *json) {
 
 	mutex_unlock(&media->ssrc_hash_out.lock);
 
+#ifdef WITH_TRANSCODING
+	struct media_player *mp = media->players[MP_DEFAULT];
+	if (mp) {
+		mutex_lock(&mp->lock);
+
+		json_builder_set_member_name(json, "media_player");
+
+		json_builder_begin_object(json);
+
+		json_builder_set_member_name(json, "duration");
+		json_builder_add_int_value(json, mp->coder.duration);
+		json_builder_set_member_name(json, "repeat");
+		json_builder_add_int_value(json, mp->opts.repeat);
+		json_builder_set_member_name(json, "frame_time");
+		json_builder_add_int_value(json, mp->last_frame_ts);
+
+		if (mp->ssrc_out && mp->media) {
+			json_builder_set_member_name(json, "SSRC");
+			json_builder_begin_object(json);
+			mqtt_ssrc_stats(mp->ssrc_out, json, mp->media);
+			json_builder_end_object(json);
+		}
+
+		json_builder_end_object(json);
+
+		mutex_unlock(&mp->lock);
+	}
+#endif
+
 	mqtt_stream_stats(ps, json);
 }
 
diff --git a/daemon/recording.c b/daemon/recording.c
index 54731ccd1..ef4f50dbf 100644
--- a/daemon/recording.c
+++ b/daemon/recording.c
@@ -345,8 +345,11 @@ static void recording_update_flags(call_t *call, bool streams) {
 
 static void rec_setup_monologue(struct call_monologue *ml) {
 	recording_setup_monologue(ml);
-	if (ml->players[MP_REC]) {
-		bool ret = media_player_start(ml->players[MP_REC]);
+	__auto_type audio = ml->audio;
+	if (!audio)
+		return;
+	if (audio->players[MP_REC]) {
+		bool ret = media_player_start(audio->players[MP_REC]);
 		if (!ret)
 			ilog(LOG_WARN, "Failed to start media player for recording announcement");
 	}
diff --git a/daemon/sdp.c b/daemon/sdp.c
index 6d423774c..35670fef9 100644
--- a/daemon/sdp.c
+++ b/daemon/sdp.c
@@ -3192,7 +3192,7 @@ static void sdp_out_add_other(GString *out, struct call_monologue *monologue,
 	if (flags->loop_protect)
 		append_attr_to_gstring(out, "rtpengine", &rtpe_instance_id, flags, media->type_id);
 #ifdef WITH_TRANSCODING
-	if (monologue->players[MP_DEFAULT] && monologue->players[MP_DEFAULT]->opts.moh && rtpe_config.moh_attr_name) {
+	if (media->players[MP_DEFAULT] && media->players[MP_DEFAULT]->opts.moh && rtpe_config.moh_attr_name) {
 		append_null_attr_to_gstring(out, rtpe_config.moh_attr_name, flags, media->type_id);
 	}
 #endif
diff --git a/daemon/t38.c b/daemon/t38.c
index ac3f01c9b..9dd6c51f0 100644
--- a/daemon/t38.c
+++ b/daemon/t38.c
@@ -415,13 +415,13 @@ int t38_gateway_pair(struct call_media *t38_media, struct call_media *pcm_media,
 	if (!(tg->gw = t38_gateway_init(NULL, t38_gateway_handler, tg)))
 		goto err;
 
-	media_player_new(&tg->pcm_player, pcm_media->monologue,
+	media_player_new(&tg->pcm_player, pcm_media,
 			call_get_first_ssrc(&pcm_media->ssrc_hash_out),
 			NULL);
 	// even though we call media_player_set_media() here, we need to call it again in
 	// t38_gateway_start because our sink might not have any streams added here yet,
 	// leaving the media_player setup incomplete
-	media_player_set_media(tg->pcm_player, pcm_media);
+	media_player_set_sink(tg->pcm_player);
 	tg->pcm_player->run_func = t38_pcm_player;
 
 	// set options
@@ -476,7 +476,7 @@ void t38_gateway_start(struct t38_gateway *tg, str_case_value_ht codec_set) {
 		return;
 
 	// set up our player first
-	media_player_set_media(tg->pcm_player, tg->pcm_media);
+	media_player_set_sink(tg->pcm_player);
 	if (media_player_setup(tg->pcm_player, &tg->pcm_pt, NULL, codec_set))
 		return;
 
diff --git a/include/call.h b/include/call.h
index bc359a87e..2a8ec04fe 100644
--- a/include/call.h
+++ b/include/call.h
@@ -565,6 +565,7 @@ struct call_media {
 	struct t38_gateway	*t38_gateway;
 	struct audio_player	*audio_player;
 	struct codec_handler	*t38_handler;
+	struct media_player	*players[MP_COUNT];
 
 	unsigned int		buffer_delay;
 
@@ -632,10 +633,12 @@ struct call_monologue {
 	GHashTable		*subscribers_ht;	/* for quick lookup */
 	medias_arr		*medias;
 	media_id_ht		media_ids;
-	struct media_player	*players[MP_COUNT];
 	struct session_bandwidth sdp_session_bandwidth;
 	GString			*last_out_sdp;
 
+	// for media players
+	struct call_media	*audio;
+
 	sdp_origin		sdp_orig_in;	/* actual origin belonging to this monologue */
 	sdp_origin		sdp_orig_out;	/* previously used origin by other other side */
 
@@ -968,6 +971,7 @@ void call_media_state_machine(struct call_media *m);
 void call_media_unkernelize(struct call_media *media, const char *reason);
 void __monologue_unconfirm(struct call_monologue *monologue, const char *);
 void __media_unconfirm(struct call_media *media, const char *);
+
 __attribute__((nonnull(1)))
 /* one monologue's state from before an offer, held as a call record snapshot */
 struct call_checkpoint {
@@ -982,6 +986,10 @@ void call_checkpoint_free_all(call_t *);
 
 void update_init_monologue_subscribers(struct call_monologue *ml, enum ng_opmode opmode);
 
+__attribute__((nonnull(1)))
+void update_init_subscribers(struct call_media *media, struct stream_params *sp,
+		sdp_ng_flags *flags, enum ng_opmode opmode);
+
 int call_stream_address(GString *, struct packet_stream *ps, enum stream_address_format format,
 		const struct local_intf *ifa, bool keep_unspec);
 
diff --git a/include/media_player.h b/include/media_player.h
index 2b1b07ff1..f88dbf9bf 100644
--- a/include/media_player.h
+++ b/include/media_player.h
@@ -66,7 +66,6 @@ struct media_player {
 	mutex_t lock;
 	media_player_run_func run_func;
 	call_t *call;
-	struct call_monologue *ml;
 	struct call_media *media;
 	struct sink_handler sink;
 
@@ -122,16 +121,16 @@ struct send_timer {
 
 #define MPO(...) (media_player_opts_t){__VA_ARGS__}
 
-void media_player_new(struct media_player **, struct call_monologue *,
+void media_player_new(struct media_player **, struct call_media *,
 		struct ssrc_entry_call *prev_ssrc, media_player_opts_t *);
 bool media_player_play(struct media_player *mp, media_player_opts_t *opts);
 bool media_player_start(struct media_player *);
 long long media_player_stop(struct media_player *);
-bool media_player_is_active(struct call_monologue *);
+bool media_player_is_active(struct call_media *);
 
 int media_player_setup(struct media_player *mp, const rtp_payload_type *src_pt,
 		const rtp_payload_type *dst_pt, str_case_value_ht codec_set);
-void media_player_set_media(struct media_player *mp, struct call_media *media);
+void media_player_set_sink(struct media_player *mp);
 bool media_player_pt_match(const struct media_player *mp, const rtp_payload_type *src_pt,
 		const rtp_payload_type *dst_pt);