diff --git a/.memory/worklog.json b/.memory/worklog.json
index aed135a..71299f2 100644
--- a/.memory/worklog.json
+++ b/.memory/worklog.json
@@ -1,19 +1,5 @@
 {
   "entries": [
-    {
-      "files_changed": 2,
-      "hash": "e6b8615",
-      "message": "auto-save 2026-05-12 17:23 (~2)",
-      "ts": "2026-05-12T17:23:21+08:00",
-      "type": "commit"
-    },
-    {
-      "files_changed": 6,
-      "hash": "6a9abea",
-      "message": "auto-save 2026-05-12 17:28 (~6)",
-      "ts": "2026-05-12T17:28:54+08:00",
-      "type": "commit"
-    },
     {
       "files_changed": 1,
       "hash": "f3ec026",
@@ -3351,6 +3337,19 @@
       "type": "session-heartbeat",
       "message": "Codex 会话活跃 · 最近命令：codex · 2 项未提交变更 · 最近提交：auto-save 2026-05-14 06:16 (~5)",
       "files_changed": 2
+    },
+    {
+      "ts": "2026-05-14T06:22:32+08:00",
+      "type": "commit",
+      "message": "auto-save 2026-05-14 06:22 (~2)",
+      "hash": "2d36ada",
+      "files_changed": 2
+    },
+    {
+      "ts": "2026-05-13T22:23:14Z",
+      "type": "session-heartbeat",
+      "message": "Claude 会话活跃 · 最近命令：claude · 1 项未提交变更 · 最近提交：auto-save 2026-05-14 06:22 (~2)",
+      "files_changed": 1
     }
   ]
 }
diff --git a/api/main.py b/api/main.py
index 6fcd747..6bfccac 100644
--- a/api/main.py
+++ b/api/main.py
@@ -448,7 +448,7 @@ def run(cmd: list[str], cwd: Path | None = None) -> str:
 # ---- 启发式选帧工具 ----
 import imagehash
 import numpy as np
-from PIL import Image, ImageEnhance, ImageFilter, ImageOps
+from PIL import Image, ImageChops, ImageEnhance, ImageFilter, ImageOps
 
 
 def _sharpness_from_gray(g: np.ndarray) -> float:
@@ -592,6 +592,7 @@ def _normalize_asset_image(
     size: AssetSize,
     background: AssetBackground = "white",
     square: bool = False,
+    fill_subject: bool = False,
 ) -> tuple[int, int]:
     import io as _io
     target_w, target_h = _asset_target_size(source_path, size, square=square)
@@ -599,7 +600,25 @@ def _normalize_asset_image(
     out_path.parent.mkdir(parents=True, exist_ok=True)
     with Image.open(_io.BytesIO(img_bytes)) as raw:
         img = raw.convert("RGB")
-        img.thumbnail((target_w, target_h), Image.Resampling.LANCZOS)
+        if fill_subject:
+            diff = ImageChops.difference(img, Image.new("RGB", img.size, bg))
+            mask = diff.convert("L").point(lambda px: 255 if px > 18 else 0)
+            bbox = mask.getbbox()
+            if bbox:
+                left, top, right, bottom = bbox
+                pad_x = round((right - left) * 0.06)
+                pad_y = round((bottom - top) * 0.06)
+                img = img.crop((
+                    max(0, left - pad_x),
+                    max(0, top - pad_y),
+                    min(img.width, right + pad_x),
+                    min(img.height, bottom + pad_y),
+                ))
+            max_w = max(1, round(target_w * 0.92))
+            max_h = max(1, round(target_h * 0.94))
+            img.thumbnail((max_w, max_h), Image.Resampling.LANCZOS)
+        else:
+            img.thumbnail((target_w, target_h), Image.Resampling.LANCZOS)
         canvas = Image.new("RGB", (target_w, target_h), bg)
         canvas.paste(img, ((target_w - img.width) // 2, (target_h - img.height) // 2))
         canvas.save(out_path, "JPEG", quality=95)
@@ -699,6 +718,8 @@ SUBJECT_VIEW_LABELS: dict[str, str] = {
     "back": "背面",
     "left": "左侧",
     "right": "右侧",
+    "three_quarter_left": "左前 45°",
+    "three_quarter_right": "右前 45°",
     "side": "侧面",
     "side_walk": "侧面走路",
     "top": "顶部视角",
@@ -727,10 +748,10 @@ def _subject_view_labels(kind: SubjectKind, requested: list[str] | None = None)
         return [
             ("front", "正面站立"),
             ("back", "背面站立"),
-            ("side", "侧面站立"),
-            ("side_walk", "侧面走路"),
-            ("expression_neutral", "中性表情"),
-            ("expression_relaxed", "放松表情"),
+            ("left", "左侧站立"),
+            ("right", "右侧站立"),
+            ("three_quarter_left", "左前 45° 站立"),
+            ("three_quarter_right", "右前 45° 站立"),
         ]
     return [
         ("front", "正面"),
diff --git a/web/components/lightbox.tsx b/web/components/lightbox.tsx
index 087944a..cfe44cd 100644
--- a/web/components/lightbox.tsx
+++ b/web/components/lightbox.tsx
@@ -36,15 +36,10 @@ const OBJECT_VIEW_OPTIONS = [
 const LIVING_VIEW_OPTIONS = [
   ["front", "正面"],
   ["back", "背面"],
-  ["side", "侧面"],
-  ["side_walk", "走路"],
-  ["expression_happy", "喜"],
-  ["expression_angry", "怒"],
-  ["expression_sad", "哀"],
-  ["expression_relaxed", "乐/放松"],
-  ["action_sit", "坐"],
-  ["action_hold", "手持"],
-  ["action_use", "使用"],
+  ["left", "左侧"],
+  ["right", "右侧"],
+  ["three_quarter_left", "左前 45°"],
+  ["three_quarter_right", "右前 45°"],
 ]
 
 type LightboxTab = "clean" | "scene" | "subject" | "review"
@@ -266,7 +261,7 @@ export function FrameLightbox({ jobId, frames, activeIndex, selected, onClose, o
         views,
       })
       onJobUpdate?.(updated)
-      toast.success(`统一主体资产包已生成 · ${views.length} 张 · ${subjectReferenceFrameIndices.length} 帧参考`)
+      toast.success(`统一主体重绘完成 · ${views.length} 张 · ${subjectReferenceFrameIndices.length} 帧参考`)
     } catch (e) {
       toast.error("主体资产包生成失败：" + (e instanceof Error ? e.message : String(e)))
     } finally {
@@ -924,7 +919,7 @@ export function FrameLightbox({ jobId, frames, activeIndex, selected, onClose, o
               <span className="text-[9.5px] text-white/35 font-normal ml-auto">→ {subjectReferenceLabel}</span>
             </div>
             <div className="mb-2 rounded-md border border-white/10 bg-white/[0.035] px-2.5 py-1.5 text-[10.5px] leading-relaxed text-white/45">
-              当前工作流只保留一个主体。多张关键帧用于补齐这个主体的角度、动作和表情；场景图会按每张关键帧单独生成。
+              当前工作流只保留一个主体。多张关键帧用于重绘这个主体的六张标准站立资产图；场景图会按每张关键帧单独生成。
             </div>
 
             {!hasUnifiedSubject && (
@@ -1053,7 +1048,7 @@ export function FrameLightbox({ jobId, frames, activeIndex, selected, onClose, o
 
 	                      <div className="mt-2 rounded-md border border-violet-300/15 bg-violet-500/[0.08] p-2">
 	                        <div className="mb-2 flex items-center justify-between gap-2">
-	                          <div className="text-[11px] font-semibold text-white/90">统一主体资产包</div>
+	                          <div className="text-[11px] font-semibold text-white/90">统一主体重绘图</div>
 	                          <span className="text-[9.5px] font-mono text-white/35">
 	                            {subjectReferenceLabel}
 	                          </span>
@@ -1084,7 +1079,7 @@ export function FrameLightbox({ jobId, frames, activeIndex, selected, onClose, o
 	                            onClick={() => handleGenerateSubjectPackage(e.id, frameIndex)}
 	                            disabled={isSubjectGenerating || activeViews.length === 0}
 	                            className="rounded bg-violet-500/70 px-1.5 py-1 text-[10px] font-medium text-white transition hover:bg-violet-400 disabled:cursor-wait disabled:opacity-45 inline-flex items-center justify-center gap-1"
-	                            title="用多张关键帧参考生成同一个主体的多视角 / 动作 / 表情资产"
+	                            title="用多张关键帧参考重绘同一个主体的六张标准站立图"
                           >
                             {isSubjectGenerating ? <Loader2 className="h-3 w-3 animate-spin" /> : <Wand2 className="h-3 w-3" />}
                             {isSubjectGenerating ? "生成" : "生成"}
@@ -1150,7 +1145,7 @@ export function FrameLightbox({ jobId, frames, activeIndex, selected, onClose, o
             )}
 
             <div className="mt-1.5 text-[10px] text-white/35 leading-relaxed">
-              主体候选来自识别结果；确认唯一主体、类型和视角后，使用参考帧生成统一主体资产包。
+              主体候选来自识别结果；确认唯一主体和类型后，使用参考帧重新绘制纯背景、占满画面的主体图。
             </div>
           </section>
         </div>
diff --git a/web/components/nodes/index.tsx b/web/components/nodes/index.tsx
index a0e6db0..c9c1c93 100644
--- a/web/components/nodes/index.tsx
+++ b/web/components/nodes/index.tsx
@@ -1561,7 +1561,7 @@ export function VisualLabNode({ data, selected }: any) {
             onClick={(e) => { e.stopPropagation(); openFirstFrame() }}
             disabled={!job || frames.length === 0}
             className="min-h-14 rounded-md border border-white/10 px-2 py-2 text-left transition hover:border-violet-300/50 hover:bg-violet-400/10 disabled:opacity-35"
-            title="用多张关键帧生成统一主体的多视角 / 动作 / 表情资产包"
+            title="用多张关键帧参考重绘统一主体的六张标准站立图"
           >
             <div className="mb-1 flex items-center gap-1 text-[var(--text-strong)] text-[12px] font-semibold">
               <Package className="h-3 w-3 text-violet-300" />