diff --git a/src/iiwa_voice/iiwa_voice/voice_generate.py b/src/iiwa_voice/iiwa_voice/voice_generate.py
new file mode 100644
index 0000000..5c891a5
--- /dev/null
+++ b/src/iiwa_voice/iiwa_voice/voice_generate.py
@@ -0,0 +1,95 @@
+import httpx
+import sounddevice as sd
+import numpy as np
+import io
+import wave
+import argparse
+
+ENDPOINTS = {
+    "customvoice": "http://localhost:8091",
+    "voicedesign": "http://localhost:8092",
+    "base": "http://localhost:8093",
+}
+
+def synthesize_and_play(
+    text: str,
+    voice: str = "Vivian",
+    model_type: str = "customvoice",
+    language: str = "English",
+    instruct: str = "",
+    base_url: str | None = None,
+):
+    url = (base_url or ENDPOINTS[model_type]) + "/v1/audio/speech"
+
+    model_map = {
+        "customvoice": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
+        "voicedesign": "Qwen/Qwen3-TTS-12Hz-0.6B-VoiceDesign",
+        "base": "Qwen/Qwen3-TTS-12Hz-0.6B-Base",
+    }
+
+    payload = {
+        "model": model_map[model_type],
+        "input": text,
+        "voice": voice,
+        "language": language,
+        "response_format": "wav",
+    }
+    if instruct:
+        payload["instruct"] = instruct
+
+    print(f"[→] Отправка запроса к {url} ...")
+
+    with httpx.Client(timeout=120) as client:
+        response = client.post(url, json=payload)
+        response.raise_for_status()
+        audio_bytes = response.content
+
+    # Читаем WAV из памяти — без записи на диск
+    with wave.open(io.BytesIO(audio_bytes)) as wf:
+        sample_rate = wf.getframerate()
+        n_channels = wf.getnchannels()
+        sample_width = wf.getsampwidth()   # байт на семпл
+        frames = wf.readframes(wf.getnframes())
+
+    # Конвертируем байты → numpy array
+    dtype_map = {1: np.int8, 2: np.int16, 4: np.int32}
+    dtype = dtype_map.get(sample_width, np.int16)
+    audio_np = np.frombuffer(frames, dtype=dtype)
+
+    if n_channels > 1:
+        audio_np = audio_np.reshape(-1, n_channels)
+
+    # Нормализуем до float32 [-1.0, 1.0] для sounddevice
+    audio_float = audio_np.astype(np.float32) / np.iinfo(dtype).max
+
+    duration = len(audio_float) / sample_rate
+    print(f"[♪] Воспроизведение: {duration:.1f} сек | {sample_rate} Hz | {n_channels}ch")
+
+    sd.play(audio_float, samplerate=sample_rate)
+    sd.wait()   # блокируемся до конца воспроизведения
+    print("[✓] Готово")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Qwen3-TTS клиент")
+    parser.add_argument("--text", default="Hello! This is Qwen3-TTS speaking.")
+    parser.add_argument("--voice", default="Vivian",
+                        help="Имя голоса (CustomVoice) или описание (VoiceDesign)")
+    parser.add_argument("--type", default="customvoice",
+                        choices=["customvoice", "voicedesign", "base"])
+    parser.add_argument("--language", default="English")
+    parser.add_argument("--instruct", default="",
+                        help="Дополнительная инструкция: 'speak slowly', 'angry tone' и т.п.")
+    parser.add_argument("--url", default=None,
+                        help="Переопределить базовый URL, напр. http://myserver:8091")
+    args = parser.parse_args()
+
+    synthesize_and_play(
+        text=args.text,
+        voice=args.voice,
+        model_type=args.type,
+        language=args.language,
+        instruct=args.instruct,
+        base_url=args.url,
+    )
+    
\ No newline at end of file
diff --git a/src/iiwa_voice/resource/docker-compose.yaml b/src/iiwa_voice/resource/docker-compose.yaml
index ea54921..76c0c24 100644
--- a/src/iiwa_voice/resource/docker-compose.yaml
+++ b/src/iiwa_voice/resource/docker-compose.yaml
@@ -53,8 +53,7 @@ services:
     ports:
       - "${CUSTOMVOICE_PORT:-8091}:8091"
     command: >
-      ${MODEL_SIZE:-Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice}
-      --omni
+      --model ${MODEL_SIZE:-Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice}
       --port 8091
       --gpu-memory-utilization ${GPU_MEM_UTIL:-0.90}
       --max-model-len ${MAX_MODEL_LEN:-4096}
@@ -72,8 +71,7 @@ services:
     ports:
       - "${VOICEDESIGN_PORT:-8092}:8091"
     command: >
-      ${MODEL_VOICEDESIGN:-Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign}
-      --omni
+      --model ${MODEL_VOICEDESIGN:-Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign}
       --port 8091
       --gpu-memory-utilization ${GPU_MEM_UTIL:-0.90}
       --max-model-len ${MAX_MODEL_LEN:-4096}
@@ -91,8 +89,7 @@ services:
     ports:
       - "${BASE_PORT:-8093}:8091"
     command: >
-      ${MODEL_BASE:-Qwen/Qwen3-TTS-12Hz-1.7B-Base}
-      --omni
+      --model ${MODEL_BASE:-Qwen/Qwen3-TTS-12Hz-1.7B-Base}
       --port 8091
       --gpu-memory-utilization ${GPU_MEM_UTIL:-0.90}
       --max-model-len ${MAX_MODEL_LEN:-4096}