diff --git a/music-lab/app/pipeline/cover.py b/music-lab/app/pipeline/cover.py
new file mode 100644
index 0000000..51f3394
--- /dev/null
+++ b/music-lab/app/pipeline/cover.py
@@ -0,0 +1,88 @@
+"""AI 커버 아트 생성 — DALL·E 3 / gpt-image-1 + 그라데이션 폴백."""
+import base64
+import logging
+import os
+from io import BytesIO
+
+import httpx
+from PIL import Image
+
+from . import storage
+from .gradient import make_gradient_with_title
+
+logger = logging.getLogger("music-lab.cover")
+
+DALLE_TIMEOUT_S = 90
+
+
+def _get_api_key() -> str:
+    return os.getenv("OPENAI_API_KEY", "")
+
+
+def _get_model() -> str:
+    return os.getenv("OPENAI_IMAGE_MODEL", "gpt-image-1")
+
+
+async def generate(*, pipeline_id: int, genre: str, prompt_template: str,
+                   mood: str = "", track_title: str = "", feedback: str = "") -> dict:
+    """커버 아트 생성. 성공 시 jpg 저장 + URL 반환. 실패 시 그라데이션 폴백.
+
+    반환: {"url": str, "used_fallback": bool, "error": str | None}
+    """
+    out_path = os.path.join(storage.pipeline_dir(pipeline_id), "cover.jpg")
+    used_fallback = False
+    error = None
+
+    api_key = _get_api_key()
+    model = _get_model()
+    if api_key:
+        try:
+            await _generate_with_dalle(prompt_template, mood, feedback, out_path,
+                                       api_key=api_key, model=model)
+        except (httpx.HTTPError, httpx.TimeoutException, KeyError, ValueError, OSError) as e:
+            logger.warning("DALL·E 실패 — 폴백: %s", e)
+            error = str(e)
+            used_fallback = True
+            make_gradient_with_title(genre, track_title, out_path)
+    else:
+        used_fallback = True
+        error = "OPENAI_API_KEY 미설정"
+        make_gradient_with_title(genre, track_title, out_path)
+
+    return {
+        "url": storage.media_url(pipeline_id, "cover.jpg"),
+        "used_fallback": used_fallback,
+        "error": error,
+    }
+
+
+async def _generate_with_dalle(prompt_template: str, mood: str,
+                                feedback: str, out_path: str,
+                                *, api_key: str, model: str) -> None:
+    prompt = prompt_template
+    if mood:
+        prompt = f"{prompt}, {mood} mood"
+    if feedback:
+        prompt = f"{prompt}. 추가 지시: {feedback}"
+    prompt = f"{prompt}, no text, high quality"
+
+    async with httpx.AsyncClient(timeout=DALLE_TIMEOUT_S) as client:
+        resp = await client.post(
+            "https://api.openai.com/v1/images/generations",
+            headers={"Authorization": f"Bearer {api_key}"},
+            json={"model": model, "prompt": prompt, "size": "1024x1024", "n": 1},
+        )
+        resp.raise_for_status()
+        data = resp.json()["data"][0]
+        if "url" in data:
+            img_resp = await client.get(data["url"])
+            img_resp.raise_for_status()
+            img_bytes = img_resp.content
+        elif "b64_json" in data:
+            img_bytes = base64.b64decode(data["b64_json"])
+        else:
+            raise ValueError("DALL·E response has neither url nor b64_json")
+        # PNG → JPG 변환
+        with Image.open(BytesIO(img_bytes)) as src:
+            img = src.convert("RGB")
+        img.save(out_path, "JPEG", quality=92)
diff --git a/music-lab/app/pipeline/gradient.py b/music-lab/app/pipeline/gradient.py
new file mode 100644
index 0000000..7d67cb4
--- /dev/null
+++ b/music-lab/app/pipeline/gradient.py
@@ -0,0 +1,38 @@
+"""장르별 그라데이션 배경 + 텍스트 오버레이 — cover/video 공용."""
+from PIL import Image, ImageDraw, ImageFont
+
+GENRE_COLORS = {
+    "lo-fi":   ((26, 26, 46),  (22, 33, 62)),
+    "phonk":   ((26, 10, 10),  (45, 0,  0)),
+    "ambient": ((13, 33, 55),  (10, 22, 40)),
+    "pop":     ((26, 10, 46),  (45, 27, 78)),
+    "default": ((17, 24, 39),  (31, 41, 55)),
+}
+
+
+def make_gradient_with_title(genre: str, title: str, out_path: str,
+                              size: tuple[int, int] = (1024, 1024),
+                              quality: int = 92) -> None:
+    w, h = size
+    top, bot = GENRE_COLORS.get(genre.lower(), GENRE_COLORS["default"])
+    with Image.new("RGB", (w, h)) as img:
+        px = img.load()
+        for y in range(h):
+            t = y / h
+            r = int(top[0] + (bot[0] - top[0]) * t)
+            g = int(top[1] + (bot[1] - top[1]) * t)
+            b = int(top[2] + (bot[2] - top[2]) * t)
+            for x in range(w):
+                px[x, y] = (r, g, b)
+
+        if title:
+            try:
+                font = ImageFont.truetype("/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", 64)
+            except OSError:
+                font = ImageFont.load_default()
+            draw = ImageDraw.Draw(img)
+            bbox = draw.textbbox((0, 0), title, font=font)
+            tw, th = bbox[2] - bbox[0], bbox[3] - bbox[1]
+            draw.text(((w - tw) // 2, (h - th) // 2), title, fill=(255, 255, 255), font=font)
+
+        img.save(out_path, "JPEG", quality=quality)
diff --git a/music-lab/app/pipeline/storage.py b/music-lab/app/pipeline/storage.py
new file mode 100644
index 0000000..16ad077
--- /dev/null
+++ b/music-lab/app/pipeline/storage.py
@@ -0,0 +1,15 @@
+"""파이프라인 산출물 디렉토리 관리."""
+import os
+
+VIDEO_DATA_DIR = os.getenv("VIDEO_DATA_DIR", "/app/data/videos")
+VIDEO_MEDIA_BASE = os.getenv("VIDEO_MEDIA_BASE", "/media/videos")
+
+
+def pipeline_dir(pipeline_id: int) -> str:
+    path = os.path.join(VIDEO_DATA_DIR, str(pipeline_id))
+    os.makedirs(path, exist_ok=True)
+    return path
+
+
+def media_url(pipeline_id: int, filename: str) -> str:
+    return f"{VIDEO_MEDIA_BASE}/{pipeline_id}/{filename}"
diff --git a/music-lab/pytest.ini b/music-lab/pytest.ini
index 4584de7..654b7eb 100644
--- a/music-lab/pytest.ini
+++ b/music-lab/pytest.ini
@@ -1,3 +1,4 @@
 [pytest]
 testpaths = tests
 pythonpath = .
+asyncio_mode = auto
diff --git a/music-lab/requirements.txt b/music-lab/requirements.txt
index 4ce3db6..6a8efd9 100644
--- a/music-lab/requirements.txt
+++ b/music-lab/requirements.txt
@@ -4,7 +4,10 @@ requests==2.32.3
 python-multipart==0.0.12
 mutagen==1.47.0
 anthropic>=0.40.0
+openai>=1.20.0
 Pillow>=11.0.0
 pytest>=8.0.0
+pytest-asyncio>=0.21
 httpx>=0.27.0
+respx>=0.21
 freezegun>=1.4
diff --git a/music-lab/tests/test_cover_generation.py b/music-lab/tests/test_cover_generation.py
new file mode 100644
index 0000000..d2ba335
--- /dev/null
+++ b/music-lab/tests/test_cover_generation.py
@@ -0,0 +1,93 @@
+import base64
+
+import pytest
+import respx
+from httpx import Response
+from app.pipeline import cover, storage
+
+
+# Real PNG bytes (1x1 red pixel) so PIL can open
+_TINY_PNG = bytes.fromhex(
+    "89504e470d0a1a0a0000000d49484452000000010000000108020000009077"
+    "53de0000000c4944415478da6300010000050001"
+    "0d0a2db40000000049454e44ae426082"
+)
+
+
+@pytest.fixture
+def tmp_storage(monkeypatch, tmp_path):
+    monkeypatch.setattr(storage, "VIDEO_DATA_DIR", str(tmp_path))
+    return tmp_path
+
+
+@pytest.mark.asyncio
+@respx.mock
+async def test_dalle_success_saves_jpg(tmp_storage, monkeypatch):
+    monkeypatch.setenv("OPENAI_API_KEY", "test-key")
+    image_url = "https://oaidalleapiprodscus.blob.core.windows.net/x.png"
+    respx.post("https://api.openai.com/v1/images/generations").mock(
+        return_value=Response(200, json={"data": [{"url": image_url}]})
+    )
+    respx.get(image_url).mock(return_value=Response(200, content=_TINY_PNG))
+
+    out = await cover.generate(pipeline_id=42, genre="lo-fi",
+                               prompt_template="moody anime", mood="chill",
+                               track_title="Test")
+    assert out["used_fallback"] is False
+    assert out["url"].startswith("/media/videos/42/cover")
+    assert (tmp_storage / "42" / "cover.jpg").exists()
+
+
+@pytest.mark.asyncio
+@respx.mock
+async def test_dalle_http_error_falls_back_to_gradient(tmp_storage, monkeypatch):
+    monkeypatch.setenv("OPENAI_API_KEY", "test-key")
+    respx.post("https://api.openai.com/v1/images/generations").mock(
+        return_value=Response(504)
+    )
+    out = await cover.generate(pipeline_id=43, genre="phonk",
+                               prompt_template="dark drift", mood="aggressive",
+                               track_title="Midnight Drive")
+    assert out["used_fallback"] is True
+    assert (tmp_storage / "43" / "cover.jpg").exists()
+
+
+@pytest.mark.asyncio
+async def test_no_api_key_falls_back(tmp_storage, monkeypatch):
+    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+    out = await cover.generate(pipeline_id=44, genre="ambient",
+                               prompt_template="x", mood="calm",
+                               track_title="Calm")
+    assert out["used_fallback"] is True
+
+
+@pytest.mark.asyncio
+@respx.mock
+async def test_dalle_with_feedback_appends_to_prompt(tmp_storage, monkeypatch):
+    import json as _json
+    monkeypatch.setenv("OPENAI_API_KEY", "test-key")
+    captured = {}
+    def hook(req):
+        captured["body"] = _json.loads(req.content)
+        return Response(200, json={"data": [{"url": "https://x"}]})
+    respx.post("https://api.openai.com/v1/images/generations").mock(side_effect=hook)
+    respx.get("https://x").mock(return_value=Response(200, content=_TINY_PNG))
+    out = await cover.generate(pipeline_id=45, genre="lo-fi",
+                          prompt_template="moody anime", mood="chill",
+                          track_title="X", feedback="더 어둡게")
+    assert "더 어둡게" in captured["body"]["prompt"]
+    assert out["used_fallback"] is False
+
+
+@pytest.mark.asyncio
+@respx.mock
+async def test_dalle_b64_response_handled(tmp_storage, monkeypatch):
+    monkeypatch.setenv("OPENAI_API_KEY", "test-key")
+    b64 = base64.b64encode(_TINY_PNG).decode()
+    respx.post("https://api.openai.com/v1/images/generations").mock(
+        return_value=Response(200, json={"data": [{"b64_json": b64}]})
+    )
+    out = await cover.generate(pipeline_id=46, genre="lo-fi",
+                                prompt_template="x", mood="", track_title="X")
+    assert out["used_fallback"] is False
+    assert (tmp_storage / "46" / "cover.jpg").exists()