diff --git a/imagecreator/backend/app/services/llm.py b/imagecreator/backend/app/services/llm.py
index 7ec364a..e7ed806 100644
--- a/imagecreator/backend/app/services/llm.py
+++ b/imagecreator/backend/app/services/llm.py
@@ -1 +1,14 @@
-"""LLM service placeholder."""
+from typing import Optional
+from google.genai import Client
+from app.config import settings
+
+class LLMService:
+    def __init__(self):
+        self.client = Client(api_key=settings.GEMINI_API_KEY)
+
+    def generate_text(self, prompt: str, system_instruction: Optional[str] = None) -> str:
+        response = self.client.models.generate_content(
+            model="gemini-2.5-flash",
+            contents=prompt
+        )
+        return response.text
\ No newline at end of file
diff --git a/imagecreator/backend/app/services/prompt.py b/imagecreator/backend/app/services/prompt.py
new file mode 100644
index 0000000..5928923
--- /dev/null
+++ b/imagecreator/backend/app/services/prompt.py
@@ -0,0 +1,28 @@
+from app.services.llm import LLMService
+
+TEXT2IMAGE_SYSTEM_PROMPT = """You are an expert at optimizing text-to-image prompts.
+Your task is to take user's simple or vague prompt and enhance it with detailed, vivid descriptions.
+Include details about: subject, setting, lighting, mood, style, composition, and quality modifiers.
+Output ONLY the optimized prompt in English, no explanations."""
+
+IMAGE2IMAGE_SYSTEM_PROMPT = """You are an expert at optimizing image-to-image prompts.
+The user wants to modify an existing image. Your task is to clearly describe what should be changed.
+
+Important rules:
+1. Identify if user wants to ADD, REMOVE, REPLACE, or ENHANCE something
+2. Be specific about what to keep vs what to change
+3. Describe the desired style, mood, and quality
+4. Output ONLY the optimized prompt in English, no explanations."""
+
+class PromptOptimizationService:
+    def __init__(self):
+        self.llm = LLMService()
+
+    def optimize_text2image(self, user_prompt: str) -> str:
+        full_prompt = f"{TEXT2IMAGE_SYSTEM_PROMPT}\n\nUser's original prompt: {user_prompt}"
+        return self.llm.generate_text(full_prompt)
+
+    def optimize_image2image(self, user_prompt: str, original_description: str = "") -> str:
+        context = f"Original image description: {original_description}\n\n" if original_description else ""
+        full_prompt = f"{IMAGE2IMAGE_SYSTEM_PROMPT}\n\n{context}User's modification request: {user_prompt}"
+        return self.llm.generate_text(full_prompt)
\ No newline at end of file
diff --git a/imagecreator/backend/tests/test_prompt.py b/imagecreator/backend/tests/test_prompt.py
new file mode 100644
index 0000000..e4569a9
--- /dev/null
+++ b/imagecreator/backend/tests/test_prompt.py
@@ -0,0 +1,17 @@
+import pytest
+from app.services.prompt import PromptOptimizationService
+
+def test_optimize_text2image_prompt():
+    service = PromptOptimizationService()
+    result = service.optimize_text2image("a cat")
+
+    assert result is not None
+    assert len(result) > len("a cat")
+    assert "cat" in result.lower()
+
+def test_optimize_image2image_prompt():
+    service = PromptOptimizationService()
+    result = service.optimize_image2image("make it colorful", "a cat image")
+
+    assert result is not None
+    assert len(result) > 0
\ No newline at end of file