sip-lab 1.20.0 → 1.22.0

package/README.md

@@ -7,13 +7,16 @@ It uses pjproject for SIP and media processing.
 
 It permits to:
   - make audio calls using UDP, TCP and TLS transports
-  - send and receive DTMF inband/RFC2833/INFO.
+  - send/receive DTMF inband/RFC2833/INFO.
   - play/record wav file on a call
   - send/receive fax (T.30 only)
   - send/receive MRCPv2 messages (TCP only, no TLS)
+  - send/receive audio using SRTP
+  - generate speech from text into a call (using flite)
 
 TODO:
   - add support for video playing/recording from/to file
+  - add support for speech recognition using pocketsphinx
   - add support for T.38 fax
   - add support for WebSocket
   - add support for WebRTC
@@ -26,7 +29,7 @@ It is distributed with prebuild binaries for node.js 15.0.0 and above (but built
 
 To install it, first install build dependencies:
 ```
-apt install build-essential automake autoconf libtool libspeex-dev libopus-dev libsdl2-dev libavdevice-dev libswscale-dev libv4l-dev libopencore-amrnb-dev libopencore-amrwb-dev libvo-amrwbenc-dev libvo-amrwbenc-dev libboost-dev libtiff-dev libpcap-dev libssl-dev uuid-dev cmake
+apt install build-essential automake autoconf libtool libspeex-dev libopus-dev libsdl2-dev libavdevice-dev libswscale-dev libv4l-dev libopencore-amrnb-dev libopencore-amrwb-dev libvo-amrwbenc-dev libvo-amrwbenc-dev libboost-dev libtiff-dev libpcap-dev libssl-dev uuid-dev flite-dev cmake
 ```
 
 Then install sip-lab (local build of the addon might be triggered here if this is not Debian 11):

package/binding.gyp

@@ -80,6 +80,12 @@
                 '-lswscale',
                 '-lavutil',
                 '-lspeex',
+		'-lflite',
+		'-lflite_cmu_us_awb',
+		'-lflite_cmu_us_kal',
+		'-lflite_cmu_us_rms',
+		'-lflite_cmu_us_slt',
+		'-lflite_cmu_us_kal16',
                 '-l srtp-x86_64-unknown-linux-gnu',
               ],
           },
@@ -109,6 +115,7 @@
         'src/addon.cpp',
         'src/pjmedia/src/pjmedia/dtmfdet.c',
         'src/pjmedia/src/pjmedia/fax_port.c',
+        'src/pjmedia/src/pjmedia/flite_port.c',
       ],
     },
   ],

package/index.js

@@ -66,6 +66,7 @@ addon.call = {
   stop_play_wav: (c_id, params) => { return addon.call_stop_play_wav(c_id, JSON.stringify(params ? params : {})) },
   start_fax: (c_id, params) => { return addon.call_start_fax(c_id, JSON.stringify(params)) },
   stop_fax: (c_id, params) => { return addon.call_stop_fax(c_id, JSON.stringify(params ? params : {})) },
+  start_speech_synth: (c_id, params) => { return addon.call_start_speech_synth(c_id, JSON.stringify(params)) },
   get_stream_stat: (c_id, params) => { return addon.call_get_stream_stat(c_id, JSON.stringify(params ? params : {})) },
   //refer: (c_id, params) => { return addon.call_refer(c_id, JSON.stringify(params)) },
   get_info: addon.call_get_info,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "sip-lab",
-  "version": "1.20.0",
+  "version": "1.22.0",
   "description": "",
   "main": "index.js",
   "engines": {
@@ -21,7 +21,7 @@
   "gypfile": true,
   "homepage": "https://github.com/MayamaTakeshi/sip-lab",
   "dependencies": {
-    "@mayama/zeq": "^4.16.0",
+    "@mayama/zeq": "^4.17.0",
     "mrcp": "^1.4.0",
     "mrcp-matching": "^1.0.0",
     "node-addon-api": "^5.0.0",

package/samples/text_to_speech.js

@@ -0,0 +1,176 @@
+var sip = require ('../index.js')
+var Zeq = require('@mayama/zeq')
+var z = new Zeq()
+var m = require('data-matching')
+var sip_msg = require('sip-matching')
+var sdp = require('sdp-matching')
+
+async function test() {
+    //sip.set_log_level(6)
+    sip.dtmf_aggregation_on(500)
+
+    z.trap_events(sip.event_source, 'event', (evt) => {
+        var e = evt.args[0]
+        return e
+    })
+
+    console.log(sip.start((data) => { console.log(data)} ))
+
+    t1 = sip.transport.create({address: "127.0.0.1", type: 'tcp'})
+    t2 = sip.transport.create({address: "127.0.0.1", type: 'tcp'})
+
+    console.log("t1", t1)
+    console.log("t2", t2)
+
+    oc = sip.call.create(t1.id, {
+        from_uri: '"abc"<sip:alice@test.com>',
+        to_uri: `sip:bob@${t2.address}:${t2.port}`,
+        headers: {
+            'X-MyHeader1': 'abc',
+            'X-MyHeader2': 'def',
+        },
+    })
+
+    await z.wait([
+        {
+            event: "incoming_call",
+            call_id: m.collect("call_id"),
+            msg: sip_msg({
+                $rm: 'INVITE',
+                $fU: 'alice',
+                $fd: 'test.com',
+                $tU: 'bob',
+                '$hdr(X-MyHeader1)': 'abc',
+                'hdr_x_myheader2': 'def',
+            }),
+        },
+        {
+            event: 'response',
+            call_id: oc.id,
+            method: 'INVITE',
+            msg: sip_msg({
+                $rs: '100',
+                $rr: 'Trying',
+                '$(hdrcnt(via))': 1,
+                'hdr_call_id': m.collect('sip_call_id'),
+                $fU: 'alice',
+                $fd: 'test.com',
+                $tU: 'bob',
+                '$hdr(l)': '0',
+            }),
+        },
+    ], 1000)
+
+    ic = {
+        id: z.store.call_id,
+        sip_call_id: z.store.sip_call_id,
+    }
+
+    sip.call.respond(ic.id, {
+        code: 200,
+        reason:'OK',
+        headers: {
+            'X-MyHeader3': 'ghi',
+            'X-MyHeader4': 'jkl',
+        },
+    })
+
+    await z.wait([
+        {
+            event: 'media_update',
+            call_id: oc.id,
+            status: 'ok',
+        },
+        {
+            event: 'media_update',
+            call_id: ic.id,
+            status: 'ok',
+        },
+        {
+            event: 'response',
+            call_id: oc.id,
+            method: 'INVITE',
+            msg: sip_msg({
+                $rs: '200',
+                $rr: 'OK',
+                '$(hdrcnt(v))': 1,
+                $fU: 'alice',
+                $fd: 'test.com',
+                $tU: 'bob',
+                '$hdr(content-type)': 'application/sdp',
+                $rb: '!{_}a=sendrecv',
+                '$hdr(X-MyHeader3)': 'ghi',
+                '$hdr(X-MyHeader4)': 'jkl',
+            }),
+        },
+    ], 1000)
+
+    await z.sleep(500)
+
+    sip.call.start_record_wav(oc.id, {file: './oc.wav'})
+    sip.call.start_record_wav(ic.id, {file: './ic.wav'})
+
+    sip.call.send_dtmf(oc.id, {digits: '1234', mode: 1})
+    sip.call.send_dtmf(ic.id, {digits: '1234', mode: 1})
+
+    await z.wait([
+	{
+		event: 'dtmf',
+		call_id: ic.id,
+		digits: '1234',
+		mode: 1,
+		media_id: 0
+	},
+	{
+		event: 'dtmf',
+		call_id: oc.id,
+		digits: '1234',
+		mode: 1,
+		media_id: 0
+	},
+    ], 3000)
+
+    sip.call.start_speech_synth(oc.id, {voice: 'slt', text: 'Hello World.'})
+    sip.call.start_speech_synth(ic.id, {voice: 'kal', text: 'How are you?'})
+
+    await z.sleep(1500)
+
+    sip.call.stop_record_wav(oc.id)
+    sip.call.stop_record_wav(ic.id)
+
+    sip.call.terminate(oc.id)
+
+    await z.wait([
+        {
+            event: 'call_ended',
+            call_id: oc.id,
+        },
+        {
+            event: 'call_ended',
+            call_id: ic.id,
+        },
+        {
+            event: 'response',
+            call_id: oc.id,
+            method: 'BYE',
+            msg: sip_msg({
+                $rs: '200',
+                $rr: 'OK',
+            }),
+        },
+    ], 1000)
+
+    await z.sleep(1000)
+
+    console.log("Success")
+
+    sip.stop()
+}
+
+
+test()
+.catch(e => {
+    console.error(e)
+    process.exit(1)
+})
+

package/src/addon.cpp

@@ -569,6 +569,40 @@ Napi::Value call_start_fax(const Napi::CallbackInfo &info) {
   return env.Null();
 }
 
+Napi::Value call_start_speech_synth(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 2) {
+    Napi::Error::New(env,
+                     "Wrong number of arguments. Expected: call_id, params.")
+        .ThrowAsJavaScriptException();
+    return env.Null();
+  }
+
+  if (!info[0].IsNumber()) {
+    Napi::TypeError::New(env, "call_id must be number.")
+        .ThrowAsJavaScriptException();
+    return env.Null();
+  }
+  int call_id = info[0].As<Napi::Number>().Int32Value();
+
+  if (!info[1].IsString()) {
+    Napi::TypeError::New(env, "params must be a JSON string.")
+        .ThrowAsJavaScriptException();
+    return env.Null();
+  }
+  const string json = info[1].As<Napi::String>().Utf8Value();
+
+  int res = pjw_call_start_speech_synth(call_id, json.c_str());
+
+  if (res != 0) {
+    Napi::Error::New(env, pjw_get_error()).ThrowAsJavaScriptException();
+    return env.Null();
+  }
+
+  return env.Null();
+}
+
 Napi::Value call_stop_record_wav(const Napi::CallbackInfo &info) {
   Napi::Env env = info.Env();
 
@@ -1265,6 +1299,9 @@ Napi::Object init(Napi::Env env, Napi::Object exports) {
   exports.Set("call_start_play_wav",
               Napi::Function::New(env, call_start_play_wav));
   exports.Set("call_start_fax", Napi::Function::New(env, call_start_fax));
+
+  exports.Set("call_start_speech_synth", Napi::Function::New(env, call_start_speech_synth));
+
   exports.Set("call_stop_record_wav",
               Napi::Function::New(env, call_stop_record_wav));
   exports.Set("call_stop_play_wav",

package/src/pjmedia/include/pjmedia/flite_port.h

@@ -0,0 +1,26 @@
+#ifndef __FLITE_PORT_H__
+#define __FLITE_PORT_H__
+
+#include <pjmedia/port.h>
+
+PJ_BEGIN_DECL
+
+PJ_DEF(pj_status_t) pjmedia_flite_port_create( pj_pool_t *pool,
+				unsigned clock_rate,
+				unsigned channel_count,
+				unsigned samples_per_frame,
+				unsigned bits_per_sample,
+				void (*cb)(pjmedia_port*, 
+					void *user_data, 
+					int result), 
+				void *user_data,
+				const char *voice,
+				pjmedia_port **p_port);
+
+PJ_DEF(pj_status_t) pjmedia_flite_port_speak( pjmedia_port *port,
+                                          const char *text,
+                                          unsigned options);
+
+PJ_END_DECL
+
+#endif	/* __FLITE_PORT_H__ */

package/src/pjmedia/src/pjmedia/flite_port.c

@@ -0,0 +1,194 @@
+/* adapted from https://github.com/signalwire/freeswitch/blob/master/src/mod/asr_tts/mod_flite/mod_flite.c */
+
+#include "flite_port.h"
+#include <pjmedia/errno.h>
+#include <pjmedia/port.h>
+#include <pj/assert.h>
+#include <pj/lock.h>
+#include <pj/pool.h>
+#include <pj/string.h>
+
+#include "siplab_constants.h"
+
+#include <flite/flite.h>
+
+#define SIGNATURE   PJMEDIA_SIGNATURE('f', 'l', 'i', 't')
+#define THIS_FILE   "flite_port.c"
+
+#if 0
+#   define TRACE_(expr)	PJ_LOG(4,expr)
+#else
+#   define TRACE_(expr)
+#endif
+
+static pj_status_t flite_get_frame(pjmedia_port *port, 
+					pjmedia_frame *frame);
+static pj_status_t flite_on_destroy(pjmedia_port *port);
+
+
+cst_voice *register_cmu_us_awb(void);
+void unregister_cmu_us_awb(cst_voice * v);
+
+cst_voice *register_cmu_us_kal(void);
+void unregister_cmu_us_kal(cst_voice * v);
+
+cst_voice *register_cmu_us_rms(void);
+void unregister_cmu_us_rms(cst_voice * v);
+
+cst_voice *register_cmu_us_slt(void);
+void unregister_cmu_us_slt(cst_voice * v);
+
+cst_voice *register_cmu_us_kal16(void);
+void unregister_cmu_us_kal16(cst_voice * v);
+
+static int initialized = 0;
+
+static struct {
+	cst_voice *awb;
+	cst_voice *kal;
+	cst_voice *rms;
+	cst_voice *slt;
+	cst_voice *kal16;
+} globals;
+
+struct flite_t {
+    struct pjmedia_port base;
+	void (*flite_cb)(pjmedia_port*, void*, int);
+	void *flite_cb_user_data;
+	cst_voice *v;
+    int written_samples;
+	cst_wave *w;
+	char *buffer;
+};
+
+#define free_wave(w) if (w) {delete_wave(w) ; w = NULL; }
+#define FLITE_BLOCK_SIZE 1024 * 32
+
+PJ_DEF(pj_status_t) pjmedia_flite_port_create( pj_pool_t *pool,
+				unsigned clock_rate,
+				unsigned channel_count,
+				unsigned samples_per_frame,
+				unsigned bits_per_sample,
+				void (*cb)(pjmedia_port*, 
+					void *user_data, 
+					int result), 
+				void *user_data,
+				const char *voice,
+				pjmedia_port **p_port)
+{
+    struct flite_t *flite;
+    const pj_str_t name = pj_str("flite_data");
+
+    if(!initialized) {
+        flite_init();
+        globals.awb = register_cmu_us_awb();
+        globals.kal = register_cmu_us_kal();
+        globals.rms = register_cmu_us_rms();
+        globals.slt = register_cmu_us_slt();
+        globals.kal16 = register_cmu_us_kal16();
+        initialized = 1;
+    }
+
+    PJ_ASSERT_RETURN(pool && clock_rate && channel_count && 
+		     samples_per_frame && bits_per_sample == 16 && 
+		     p_port != NULL, PJ_EINVAL);
+
+    PJ_ASSERT_RETURN(pool && p_port, PJ_EINVAL);
+
+    flite = PJ_POOL_ZALLOC_T(pool, struct flite_t);
+    PJ_ASSERT_RETURN(pool != NULL, PJ_ENOMEM);
+
+    pjmedia_port_info_init(&flite->base.info, &name, SIGNATURE, clock_rate,
+			   channel_count, bits_per_sample, samples_per_frame);
+
+    flite->base.get_frame = &flite_get_frame;
+    flite->base.on_destroy = &flite_on_destroy;
+
+	if (!strcasecmp(voice, "awb")) {
+		flite->v = globals.awb;
+	} else if (!strcasecmp(voice, "kal")) {
+/*  "kal" is 8kHz and the native rate is set to 16kHz
+ *  so kal talks a little bit too fast ...
+ *  for now: "symlink" kal to kal16
+ */		flite->v = globals.kal16;
+	} else if (!strcasecmp(voice, "rms")) {
+		flite->v = globals.rms;
+	} else if (!strcasecmp(voice, "slt")) {
+		flite->v = globals.slt;
+	} else if (!strcasecmp(voice, "kal16")) {
+		flite->v = globals.kal16;
+	} else {
+		TRACE_((THIS_FILE, "Invalid voice"));
+        return 0;
+	}
+
+    flite->flite_cb = cb;
+    flite->flite_cb_user_data = user_data;
+
+    TRACE_((THIS_FILE, "flite_device created: %u/%u/%u/%u", clock_rate, 
+	    channel_count, samples_per_frame, bits_per_sample));
+
+    *p_port = &flite->base; 
+    return PJ_SUCCESS;
+}
+
+PJ_DEF(pj_status_t) pjmedia_flite_port_speak( pjmedia_port *port,
+                                          const char *text,
+                                          unsigned options) {
+    struct flite_t *flite = (struct flite_t*)port;
+    if(flite->w) {
+        free_wave(flite->w);
+    }
+    
+    flite->w = flite_text_to_wave(text, flite->v);
+    if (flite->w->sample_rate != PJMEDIA_PIA_SRATE(&port->info)) {
+		cst_wave_resample(flite->w, PJMEDIA_PIA_SRATE(&port->info));
+    }
+    flite->written_samples = 0;
+
+    return PJ_SUCCESS;
+}
+
+// called when pjmedia needs data to be sent out
+static pj_status_t flite_get_frame(pjmedia_port *port,
+					pjmedia_frame *frame) {
+
+	PJ_ASSERT_RETURN(port && frame, PJ_EINVAL);
+
+    struct flite_t *flite = (struct flite_t*)port;
+
+    if(!flite->w) {
+        printf("flite no data\n");
+        frame->type = PJMEDIA_FRAME_TYPE_NONE;
+        return PJ_SUCCESS;
+    }
+
+    printf("written_samples=%i num_samples=%i\n", flite->written_samples, flite->w->num_samples);
+    if (flite->written_samples + PJMEDIA_PIA_SPF(&port->info) > flite->w->num_samples) {
+        printf("flite no more data\n");
+        free_wave(flite->w);
+        frame->type = PJMEDIA_FRAME_TYPE_NONE;
+        return PJ_SUCCESS;
+    }
+
+    memcpy(frame->buf, flite->w->samples + flite->written_samples, PJMEDIA_PIA_SPF(&port->info)*2);
+    flite->written_samples += PJMEDIA_PIA_SPF(&port->info);
+    frame->type = PJMEDIA_FRAME_TYPE_AUDIO;
+    printf("flite data written samples=%i\n", PJMEDIA_PIA_SPF(&port->info));
+
+    return PJ_SUCCESS;
+}
+
+/*
+ * Destroy port.
+ */
+static pj_status_t flite_on_destroy(pjmedia_port *port)
+{
+	printf("flite_on_destroy\n");
+
+    struct flite_t *flite = (struct flite_t*)port;
+    free_wave(flite->w);
+
+    return PJ_SUCCESS;
+}
+

package/src/sip.cpp

@@ -22,6 +22,7 @@
 
 #include "dtmfdet.h"
 #include "fax_port.h"
+#include "flite_port.h"
 
 #include <ctime>
 
@@ -329,6 +330,7 @@ struct AudioEndpoint {
   ConfBridgePort tonegen_cbp;
   ConfBridgePort dtmfdet_cbp;
   ConfBridgePort fax_cbp;
+  ConfBridgePort flite_cbp;
 };
 
 struct VideoEndpoint {
@@ -620,6 +622,7 @@ bool prepare_wav_player(Call *c, AudioEndpoint *ae, const char *file);
 bool prepare_wav_writer(Call *c, AudioEndpoint *ae, const char *file);
 bool prepare_fax(Call *c, AudioEndpoint *ae, bool is_sender, const char *file,
                  unsigned flags);
+bool prepare_flite(Call *c, AudioEndpoint *ae, const char *voice);
 
 void prepare_error_event(ostringstream *oss, char *scope, char *details);
 // void prepare_pjsipcall_error_event(ostringstream *oss, char *scope, char
@@ -638,6 +641,7 @@ typedef pj_status_t (*audio_endpoint_stop_op_t)(Call *call, AudioEndpoint *ae);
 pj_status_t audio_endpoint_stop_play_wav(Call *call, AudioEndpoint *ae);
 pj_status_t audio_endpoint_stop_record_wav(Call *call, AudioEndpoint *ae);
 pj_status_t audio_endpoint_stop_fax(Call *call, AudioEndpoint *ae);
+pj_status_t audio_endpoint_stop_flite(Call *call, AudioEndpoint *ae);
 
 static pjsip_module mod_tester = {
     NULL,
@@ -3176,6 +3180,37 @@ out:
   return 0;
 }
 
+pj_status_t audio_endpoint_remove_port(Call *call, ConfBridgePort *cbp) {
+  printf("audio_endpoint_remove_port\n");
+  pj_status_t status;
+
+  if(cbp->port) {
+    /* 
+    no need to call pjmedia_conf_disconnect_port because pjmedia_conf_remove_port calls:
+      pjmedia_conf_disconnect_port_from_sources(conf, port);
+      pjmedia_conf_disconnect_port_from_sinks(conf, port);
+    */
+
+    status = pjmedia_conf_remove_port(call->conf, cbp->slot);
+    if (status != PJ_SUCCESS) {
+      set_error("pjmedia_conf_remove_port failed");
+      return false;
+    }
+    cbp->slot = 0;
+
+    status = pjmedia_port_destroy(cbp->port);
+    if (status != PJ_SUCCESS) {
+      set_error("pjmedia_port_destroy failed");
+      return false;
+    }
+    cbp->port = NULL;
+  }
+
+  printf("success\n");
+  return PJ_SUCCESS;
+}
+
+
 int pjw_call_reinvite(long call_id, const char *json) {
   addon_log(L_DBG, "pjw_call_reinvite call_id=%d\n", call_id);
 
@@ -3609,6 +3644,125 @@ out:
   return 0;
 }
 
+pj_status_t audio_endpoint_start_speech_synth(Call *call, AudioEndpoint *ae, const char * voice, const char *text) {
+  pj_status_t status;
+
+  if(!ae->stream_cbp.port) {
+    set_error("stream port is not ready yet");
+    return -1;
+  }
+
+  // First stop and destroy existing flite port.
+  status = audio_endpoint_stop_flite(call, ae);
+  if(status != PJ_SUCCESS) {
+    return -1;
+  }
+
+  if (!prepare_flite(call, ae, voice)) {
+    return -1;
+  }
+
+  pjmedia_flite_port_speak(ae->flite_cbp.port, text, 0);
+
+  return PJ_SUCCESS;
+}
+
+int pjw_call_start_speech_synth(long call_id, const char *json) {
+  PJW_LOCK();
+  clear_error();
+
+  long val;
+  Call *call;
+
+  MediaEndpoint *me;
+  AudioEndpoint *ae;
+  int ae_count;
+
+  unsigned media_id = 0;
+
+  char *voice;
+
+  char *text;
+
+  char buffer[MAX_JSON_INPUT];
+
+  Document document;
+
+  const char *valid_params[] = {"voice", "text", "media_id", ""};
+
+  if (!g_call_ids.get(call_id, val)) {
+    set_error("Invalid call_id");
+    goto out;
+  }
+  call = (Call *)val;
+
+  ae_count = count_media_by_type(call, ENDPOINT_TYPE_AUDIO);
+
+  if (ae_count == 0) {
+    set_error("No audio endpoint");
+    goto out;
+  }
+
+  if (!parse_json(document, json, buffer, MAX_JSON_INPUT)) {
+    goto out;
+  }
+
+  if (!validate_params(document, valid_params)) {
+    goto out;
+  }
+
+  if (json_get_string_param(document, "voice", false, &voice) <= 0) {
+    goto out;
+  }
+
+  if (!voice[0]) {
+    set_error("voice cannot be blank string");
+    goto out;
+  }
+
+  if (json_get_string_param(document, "text", false, &text) <= 0) {
+    goto out;
+  }
+
+  if (!text[0]) {
+    set_error("text cannot be blank string");
+    goto out;
+  }
+
+  if (ae_count > 1) {
+    if (json_get_uint_param(document, "media_id", false, &media_id) <= 0) {
+      goto out;
+    }
+  }
+
+  if ((int)media_id >= call->media_count) {
+    set_error("invalid media_id");
+    goto out;
+  }
+
+  me = (MediaEndpoint *)call->media[media_id];
+  if (ENDPOINT_TYPE_AUDIO != me->type) {
+    set_error("media_endpoint is not audio endpoint");
+    goto out;
+  }
+
+  ae = (AudioEndpoint *)me->endpoint.audio;
+
+  audio_endpoint_start_speech_synth(call, ae, voice, text);
+
+out:
+  PJW_UNLOCK();
+  if (pjw_errorstring[0]) {
+    return -1;
+  }
+
+  return 0;
+}
+
+pj_status_t audio_endpoint_stop_flite(Call *call, AudioEndpoint *ae) {
+  return audio_endpoint_remove_port(call, &ae->flite_cbp);
+}
+
 pj_status_t call_stop_audio_endpoints_op(Call *call,
                                          audio_endpoint_stop_op_t op) {
   addon_log(L_DBG, "call_stop_audio_endpoints_op media_count=%d\n",
@@ -3630,37 +3784,6 @@ pj_status_t call_stop_audio_endpoints_op(Call *call,
   return PJ_SUCCESS;
 }
 
-pj_status_t audio_endpoint_remove_port(Call *call, ConfBridgePort *cbp) {
-  printf("audio_endpoint_remove_port\n");
-  pj_status_t status;
-
-  if(cbp->port) {
-    /* 
-    no need to call pjmedia_conf_disconnect_port because pjmedia_conf_remove_port calls:
-      pjmedia_conf_disconnect_port_from_sources(conf, port);
-      pjmedia_conf_disconnect_port_from_sinks(conf, port);
-    */
-
-    status = pjmedia_conf_remove_port(call->conf, cbp->slot);
-    if (status != PJ_SUCCESS) {
-      set_error("pjmedia_conf_remove_port failed");
-      return false;
-    }
-    cbp->slot = 0;
-
-    status = pjmedia_port_destroy(cbp->port);
-    if (status != PJ_SUCCESS) {
-      set_error("pjmedia_port_destroy failed");
-      return false;
-    }
-    cbp->port = NULL;
-  }
-
-  printf("success\n");
-  return PJ_SUCCESS;
-}
-
-
 pj_status_t audio_endpoint_stop_play_wav(Call *call, AudioEndpoint *ae) {
   return audio_endpoint_remove_port(call, &ae->wav_player_cbp);
 }
@@ -6394,7 +6517,6 @@ bool prepare_tonegen(Call *c, AudioEndpoint *ae) {
     return true;
   }
 
-  printf("call.id=%i p1\n", c->id);
   status = pjmedia_tonegen_create(
         c->inv->pool, PJMEDIA_PIA_SRATE(&ae->stream_cbp.port->info),
         PJMEDIA_PIA_CCNT(&ae->stream_cbp.port->info),
@@ -6405,21 +6527,18 @@ bool prepare_tonegen(Call *c, AudioEndpoint *ae) {
     return false;
   }
 
-  printf("call.id=%i p2\n", c->id);
   status = pjmedia_conf_add_port(c->conf, c->inv->pool, ae->tonegen_cbp.port, NULL, &ae->tonegen_cbp.slot);
   if (status != PJ_SUCCESS) {
     set_error("pjmedia_conf_add_port failed");
     return false;
   }
 
-  printf("call.id=%i p3\n", c->id);
   status = pjmedia_conf_connect_port(c->conf, ae->tonegen_cbp.slot, ae->stream_cbp.slot, 0);
   if (status != PJ_SUCCESS) {
     set_error("pjmedia_conf_connect_port failed");
     return false;
   }
 
-  printf("call.id=%i p4 (success)\n", c->id);
   return true;
 }
 
@@ -6553,6 +6672,41 @@ bool prepare_fax(Call *c, AudioEndpoint *ae, bool is_sender, const char *file,
   return true;
 }
 
+bool prepare_flite(Call *c, AudioEndpoint *ae, const char *voice) {
+  printf("prepare_flite call.id=%i\n", c->id);
+  pj_status_t status;
+
+  if(ae->flite_cbp.port) {
+    printf("already prepared\n");
+    return true;
+  }
+
+  status = pjmedia_flite_port_create(
+        c->inv->pool, PJMEDIA_PIA_SRATE(&ae->stream_cbp.port->info),
+        PJMEDIA_PIA_CCNT(&ae->stream_cbp.port->info),
+        PJMEDIA_PIA_SPF(&ae->stream_cbp.port->info),
+        PJMEDIA_PIA_BITS(&ae->stream_cbp.port->info), NULL, 0, voice, &ae->flite_cbp.port);
+  if (status != PJ_SUCCESS) {
+    set_error("pjmedia_flite_port_create failed");
+    return false;
+  }
+
+  status = pjmedia_conf_add_port(c->conf, c->inv->pool, ae->flite_cbp.port, NULL, &ae->flite_cbp.slot);
+  if (status != PJ_SUCCESS) {
+    set_error("pjmedia_conf_add_port failed");
+    return false;
+  }
+
+  status = pjmedia_conf_connect_port(c->conf, ae->flite_cbp.slot, ae->stream_cbp.slot, 0);
+  if (status != PJ_SUCCESS) {
+    set_error("pjmedia_conf_connect_port failed");
+    return false;
+  }
+
+  return true;
+}
+
+
 void on_rx_notify(pjsip_evsub *sub, pjsip_rx_data *rdata, int *p_st_code,
                   pj_str_t **p_st_text, pjsip_hdr *res_hdr,
                   pjsip_msg_body **p_body) {

package/src/sip.hpp

@@ -57,6 +57,8 @@ int pjw_call_start_fax(long call_id, const char *json);
 
 int pjw_call_stop_fax(long call_id, const char *json);
 
+int pjw_call_start_speech_synth(long call_id, const char *json);
+
 int pjw_call_get_stream_stat(long call_id, const char *json, char *out_stats);
 
 // int pjw_call_refer(long call_id, const char *json, long