diff --git a/backend/speech_recognition.py b/backend/speech_recognition.py
new file mode 100644
index 000000000..f740bc7ac
--- /dev/null
+++ b/backend/speech_recognition.py
@@ -0,0 +1,35 @@
+import subprocess
+import sys
+import winsound
+import whisper
+
+
+def generate_test_audio(text, output_path="test.wav"):
+    # Run pyttsx3 in a subprocess to avoid COM/torch conflict
+    script = (
+        f"import pyttsx3; e = pyttsx3.init(); "
+        f"e.save_to_file({repr(text)}, {repr(output_path)}); e.runAndWait()"
+    )
+    subprocess.run([sys.executable, "-c", script], check=True)
+    print(f"Test audio saved to: {output_path}")
+    return output_path
+
+
+def recognize_speech(wav_file):
+    model = whisper.load_model("base")
+    result = model.transcribe(wav_file)
+    return result["text"]
+
+
+def play_audio(audio_path):
+    winsound.PlaySound(audio_path, winsound.SND_FILENAME)
+
+
+if __name__ == "__main__":
+    test_text = "Hello, this is a test of the speech recognition system."
+    wav_file = generate_test_audio(test_text)
+
+    transcription = recognize_speech(wav_file)
+    print("Recognized:", transcription)
+
+    play_audio(wav_file)