diff --git a/jarvis/utils/client_utils.py b/jarvis/utils/client_utils.py
index bbe854c..cd4837d 100644
--- a/jarvis/utils/client_utils.py
+++ b/jarvis/utils/client_utils.py
@@ -1,15 +1,38 @@
-import os
+import requests
+from requests.structures import CaseInsensitiveDict
 
 from jarvis.utils import config_utils
 
 
-def speak(sentence, quality=config_utils.get_in_config("LARYNX_QUALITY")):
-    voice = config_utils.get_in_config("LARYNX_VOICE")
-    denoiserstrength = config_utils.get_in_config("LARYNX_DENOISER_STRENGTH")
+def speak(sentence, client_ip, client_port):
+    raw_audio_bytes = get_audio_from_sentence(sentence)
+    play_audio_on_client(raw_audio_bytes, client_ip, client_port)
 
-    os.system(
-        "larynx \"" + sentence + "\" --voice " + voice + " --quality " + quality +
-        # " --output-dir wav " +
-        " --interactive " +
-        "--denoiser-strength " + str(denoiserstrength) +
-        " --output-naming time")
+
+def get_audio_from_sentence(sentence):
+    voice = config_utils.get_in_config("TTS_VOICE")
+
+    headers = {'accept': '*/*'}
+
+    params = (
+        ('voice', voice),
+        ('text', sentence),
+    )
+
+    # TODO : add support for external opentts server
+    response = requests.get('http://localhost:5500/api/tts', headers=headers, params=params)
+    return response.content
+
+
+def play_audio_on_client(raw_audio_bytes, client_ip, client_port):
+    url_service = "http://" + client_ip + ":" + client_port + "/play_raw_audio"
+
+    headers = CaseInsensitiveDict()
+    headers["Content-Type"] = "text/xml; charset=utf8"
+    # headers["Authorization"] = config_utils.get_in_config("API_KEY")
+
+    response = requests.post(url_service,
+                             headers=headers,
+                             data=raw_audio_bytes)
+
+    print(response.content)