diff --git a/core/backend/apps/ai/model_library.py b/core/backend/apps/ai/model_library.py
index 1ce85cc..24fa96d 100644
--- a/core/backend/apps/ai/model_library.py
+++ b/core/backend/apps/ai/model_library.py
@@ -13,7 +13,7 @@ import uuid
 from io import BytesIO
 
 from apps.ai.models import ModelConfig
-from apps.ai.services import _asset_preview_url, build_provider, get_default_model
+from apps.ai.services import _asset_preview_url, build_person_frontal_prompt, build_provider, get_default_model
 from apps.assets.models import Asset, AssetFile
 from apps.assets.storage import TosStorage
 
@@ -67,8 +67,8 @@ def generate_model(*, team, user, brief: str, name: str | None = None) -> dict:
         raise ValueError(f"当前图像模型 {model_config.provider.name}:{model_config.name} 不支持参考图三视图(image_edit)")
 
     label = name or brief[:16]
-    # 1) 9:16 氛围正面图
-    frontal_prompt = f"{brief}，电商真人模特,9:16竖屏氛围正面半身,自然妆容,柔和影棚光,真实质感,单人,简洁背景"
+    # 1) 9:16 氛围正面图(画幅由 size 控制;提示词用统一的人物正面模板:全身 / 纯色背景)
+    frontal_prompt = build_person_frontal_prompt(brief)
     resp = provider.image_generation(model=model_config.name, prompt=frontal_prompt, size="1024x1536")
     frontal = _store_model_asset(
         team=team, user=user, media=provider.extract_first_media_url(resp),
diff --git a/core/backend/apps/ai/services.py b/core/backend/apps/ai/services.py
index b11b3e4..1f2503e 100644
--- a/core/backend/apps/ai/services.py
+++ b/core/backend/apps/ai/services.py
@@ -464,6 +464,14 @@ def build_platform_cover_prompt_refs(product, has_model: bool, base_prompt: str
     return " ".join(lines)
 
 
+def build_person_frontal_prompt(description: str = "") -> str:
+    """人物正面氛围图提示词:把脚本提取(或用户输入)的人物描述包成统一模板。
+    用户钦定格式:电商真人模特，氛围正面全身照，<描述>，自然妆容，柔和影棚光，真实质感，单人，纯色背景。"""
+    desc = (description or "").strip()
+    inner = f"{desc}，" if desc else ""
+    return f"电商真人模特，氛围正面全身照，{inner}自然妆容，柔和影棚光，真实质感，单人，纯色背景"
+
+
 def generate_base_asset(*, project, user, kind: str, prompt: str, label: str = "", group_id: str | None = None) -> AITask:
     """提交基础资产生成(**异步**):Web 请求只建 RESERVED 任务 + 预留额度(秒级),
     慢出图(文生图 / 商品 image_edit)交给 Celery worker(run_base_asset_task)跑。
@@ -480,7 +488,12 @@ def generate_base_asset(*, project, user, kind: str, prompt: str, label: str = "
     # 无主图或当前模型不支持 image_edit → 回落纯文生图(仅凭商品名脑补,不保证还原真实包装)。
     product_ref_url = _product_cover_url(project.product) if kind == BaseAssetGroup.Kind.PRODUCT else ""
     use_edit = bool(product_ref_url) and hasattr(provider, "image_edit")
-    gen_prompt = build_product_triview_prompt_refs(project.product, prompt) if use_edit else prompt
+    if use_edit:
+        gen_prompt = build_product_triview_prompt_refs(project.product, prompt)
+    elif kind == BaseAssetGroup.Kind.PERSON:
+        gen_prompt = build_person_frontal_prompt(prompt)  # 人物立绘:包成「电商真人模特/正面全身/纯色背景」统一模板
+    else:
+        gen_prompt = prompt
     payload = {
         "model": model_config.name, "endpoint": model_config.endpoint, "prompt": gen_prompt,
         "kind": kind, "label": label or "", "group_id": str(group_id) if group_id else "",