fix(core): 脚本 agent 再兼容 lines/caption 变体,补全旁白漏空

豆包不同型号吐不同 schema:Lite 用 dialogue(字符串)+subtitle;Pro 用
lines:[{role,content}]+caption。上一版只认 dialogue/subtitle,导致 Pro 变体
画面能填、旁白仍空。normalize_draft 再扩容:
- 口播来源并入 lines(每项 content/line/text)、voiceover;subtitle/caption 兜底
- 画面来源并入 visual_description
新增 lines/caption 变体回归用例,全套 22 测试通过。

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

core/backend/apps/ai/script_agent.py

@@ -288,28 +288,29 @@ def normalize_draft(raw_text: str, *, aspect_ratio: str, total_duration: int) ->
         speaker = speaker if (speaker in valid_ids) else None
         refs = [r for r in (seg.get("entity_refs") or []) if r in valid_ids]
         # 对白(剧情向):[{speaker(合法 entity id 或 null=旁白), line}];默认空 = 纯口播。
-        # 模型变体里 dialogue 可能是字符串(整句口播)而非数组——只在数组时按结构解析。
+        # 模型变体的口播字段五花八门:dialogue(字符串/数组)/ lines(数组)/ 每项 line|text|content。
+        # 这里把任意数组形态归一成结构化对白,字符串形态留给下面当整句旁白。
         raw_dialogue = seg.get("dialogue")
         dialogue = []
-        if isinstance(raw_dialogue, list):
-            for d in raw_dialogue:
-                if not isinstance(d, dict):
-                    continue
-                line = (d.get("line") or d.get("text") or "").strip()
-                if not line:
-                    continue
-                sp = d.get("speaker")
-                dialogue.append({"speaker": sp if sp in valid_ids else None, "line": line})
-        # 旁白:narration > 字符串 dialogue(整句口播)> subtitle(字幕)> 结构化对白拼接
-        narration = (seg.get("narration") or "").strip()
+        dialogue_array = raw_dialogue if isinstance(raw_dialogue, list) else (seg.get("lines") if isinstance(seg.get("lines"), list) else [])
+        for d in dialogue_array:
+            if not isinstance(d, dict):
+                continue
+            line = (d.get("line") or d.get("text") or d.get("content") or "").strip()
+            if not line:
+                continue
+            sp = d.get("speaker")
+            dialogue.append({"speaker": sp if sp in valid_ids else None, "line": line})
+        # 旁白:narration > 字符串 dialogue(整句口播)> 结构化对白/lines 拼接 > subtitle/caption(字幕兜底)
+        narration = (seg.get("narration") or seg.get("voiceover") or "").strip()
         if not narration and isinstance(raw_dialogue, str):
             narration = raw_dialogue.strip()
-        if not narration:
-            narration = (seg.get("subtitle") or "").strip()
         if not narration and dialogue:
             narration = " ".join(d["line"] for d in dialogue)  # 扁平拼接,兼容下游字幕/配音
-        # 画面:visual > visual_prompt > scene(模型变体常用 scene 描述画面)
-        visual = (seg.get("visual") or seg.get("visual_prompt") or seg.get("scene") or "").strip()
+        if not narration:
+            narration = (seg.get("subtitle") or seg.get("caption") or "").strip()
+        # 画面:visual > visual_prompt > scene > shot/visual_description(模型变体常用 scene 描述画面)
+        visual = (seg.get("visual") or seg.get("visual_prompt") or seg.get("scene") or seg.get("visual_description") or "").strip()
         norm_segments.append(
             {
                 "index": i,