diff --git a/scripts/test-xunfei-stt.py b/scripts/test-xunfei-stt.py
new file mode 100644
index 0000000..9026771
--- /dev/null
+++ b/scripts/test-xunfei-stt.py
@@ -0,0 +1,184 @@
+#!/usr/bin/env python3
+"""
+测试讯飞语音听写 WebSocket STT API
+用法: python3 test-xunfei-stt.py <APP_ID> <API_KEY> <API_SECRET> [audio_file.pcm]
+
+如果不指定音频文件，会自动生成一段 1.5 秒的 440Hz 测试音（16kHz 16bit mono PCM）
+"""
+
+import sys
+import hmac
+import hashlib
+import base64
+import json
+import time
+import struct
+import math
+import websocket
+from datetime import datetime, timezone
+from urllib.parse import urlencode
+from email.utils import formatdate
+
+def build_url(app_id, api_key, api_secret):
+    host = "iat-api.xfyun.cn"
+    path = "/v2/iat"
+    date = formatdate(usegmt=True)
+
+    signature_origin = f"host: {host}\ndate: {date}\nGET {path} HTTP/1.1"
+    signature = base64.b64encode(
+        hmac.new(api_secret.encode(), signature_origin.encode(), hashlib.sha256).digest()
+    ).decode()
+
+    authorization_origin = (
+        f'api_key="{api_key}", algorithm="hmac-sha256", '
+        f'headers="host date request-line", signature="{signature}"'
+    )
+    authorization = base64.b64encode(authorization_origin.encode()).decode()
+
+    params = urlencode({
+        "authorization": authorization,
+        "date": date,
+        "host": host,
+    })
+    return f"wss://{host}{path}?{params}"
+
+
+def generate_test_pcm(duration_sec=1.5, sample_rate=16000, freq=440):
+    """生成一段正弦波 PCM 测试音频（16bit mono）"""
+    n = int(duration_sec * sample_rate)
+    samples = []
+    for i in range(n):
+        v = math.sin(2 * math.pi * freq * i / sample_rate)
+        samples.append(int(v * 16383))  # 50% 音量
+    return struct.pack(f"<{n}h", *samples)
+
+
+def send_audio(ws, app_id, pcm_data, chunk_ms=40):
+    """按帧发送 PCM 音频数据"""
+    chunk_size = int(16000 * 2 * chunk_ms / 1000)  # 16kHz * 16bit * chunk_ms
+    total = len(pcm_data)
+    offset = 0
+    frame_idx = 0
+
+    while offset < total:
+        chunk = pcm_data[offset:offset + chunk_size]
+        offset += chunk_size
+        is_last = offset >= total
+        status = 2 if is_last else (1 if frame_idx > 0 else 0)
+
+        frame = {"data": {
+            "status": status,
+            "format": "audio/L16;rate=16000",
+            "encoding": "raw",
+            "audio": base64.b64encode(chunk).decode(),
+        }}
+
+        if frame_idx == 0:
+            frame["common"] = {"app_id": app_id}
+            frame["business"] = {
+                "language": "zh_cn",
+                "domain": "iat",
+                "accent": "mandarin",
+                "vad_eos": 3000,
+                "dwa": "wpgs",
+            }
+
+        ws.send(json.dumps(frame))
+        frame_idx += 1
+        time.sleep(chunk_ms / 1000.0)
+
+    print(f"[发送完毕] 共 {frame_idx} 帧，{total} 字节")
+
+
+def main():
+    if len(sys.argv) < 4:
+        print(__doc__)
+        sys.exit(1)
+
+    app_id = sys.argv[1]
+    api_key = sys.argv[2]
+    api_secret = sys.argv[3]
+    audio_file = sys.argv[4] if len(sys.argv) > 4 else None
+
+    if audio_file:
+        with open(audio_file, "rb") as f:
+            pcm_data = f.read()
+        print(f"[音频] 读取文件 {audio_file}，大小 {len(pcm_data)} 字节")
+    else:
+        pcm_data = generate_test_pcm(duration_sec=1.5)
+        print(f"[音频] 生成 1.5 秒 440Hz 正弦波测试音，大小 {len(pcm_data)} 字节")
+
+    url = build_url(app_id, api_key, api_secret)
+    print(f"[连接] {url[:80]}...")
+
+    results = []
+    errors = []
+    connected = [False]
+
+    def on_open(ws):
+        connected[0] = True
+        print("[WebSocket] 连接成功，开始发送音频...")
+        import threading
+        t = threading.Thread(target=send_audio, args=(ws, app_id, pcm_data))
+        t.daemon = True
+        t.start()
+
+    def on_message(ws, msg):
+        try:
+            data = json.loads(msg)
+            code = data.get("code", -1)
+            if code != 0:
+                err = f"API错误 code={code}: {data.get('message', '')}"
+                print(f"[错误] {err}")
+                errors.append(err)
+                return
+
+            ws_items = data.get("data", {}).get("result", {}).get("ws", [])
+            text = "".join(
+                cw.get("w", "")
+                for item in ws_items
+                for cw in item.get("cw", [])
+            )
+            status = data.get("data", {}).get("status", -1)
+            if text:
+                results.append(text)
+                final = "【最终】" if status == 2 else "【中间】"
+                print(f"[识别] {final} {text}")
+
+            if status == 2:
+                print("[完成] 讯飞 STT 服务正常，识别完毕")
+                ws.close()
+        except Exception as e:
+            print(f"[解析错误] {e}: {msg}")
+
+    def on_error(ws, err):
+        print(f"[WebSocket错误] {err}")
+        errors.append(str(err))
+
+    def on_close(ws, code, reason):
+        print(f"[关闭] code={code} reason={reason}")
+
+    ws = websocket.WebSocketApp(
+        url,
+        on_open=on_open,
+        on_message=on_message,
+        on_error=on_error,
+        on_close=on_close,
+    )
+
+    ws.run_forever(ping_interval=0)
+
+    print("\n===== 测试结果 =====")
+    if errors:
+        print(f"❌ 失败: {errors}")
+    elif not connected[0]:
+        print("❌ 连接失败（URL 认证错误或网络问题）")
+    elif results:
+        print(f"✅ 成功，识别文本: {''.join(results)}")
+        print("   （测试音是正弦波，无实际文字，结果为空也属正常）")
+    else:
+        print("✅ 连接和认证成功（测试音无文字内容，识别结果为空）")
+
+
+if __name__ == "__main__":
+    main()