feat: refine UI, improve UX, optimize the analysis pipeline, and add character standing positions

2026-04-02 17:39:16 +08:00
parent c3e74c228a
commit 9703714b69
153 changed files with 4472 additions and 1088 deletions
--- a/lib/prompts/novel-promotion/agent_cinematographer.en.txt
+++ b/lib/prompts/novel-promotion/agent_cinematographer.en.txt
@@ -29,5 +29,9 @@ Rules:
 2. Keep continuity across neighboring panels.
 3. Adapt to scene_type and story rhythm.
 4. Technical notes must be directly actionable by image/video generation.
-5. JSON only, no markdown.
-6. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
+5. If characters already carry `slot`, treat it as a preferred placement anchor, not an absolute boundary.
+6. When a panel is about movement, entry/exit, path traversal, transition space, temporary space, empty space, imagination, dream, memory, or abstract/non-literal space, composition and placement may deviate from a static slot if the shot logic requires it.
+7. Treat `slot` as one full placement phrase from the location context, not as a short token, whenever you reference it.
+8. Do not shorten, rewrite, summarize, or replace a provided `slot` phrase with a short token.
+9. JSON only, no markdown.
+10. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
--- a/lib/prompts/novel-promotion/agent_cinematographer.zh.txt
+++ b/lib/prompts/novel-promotion/agent_cinematographer.zh.txt
@@ -129,8 +129,11 @@
 3. 每个元素必须包含 panel_number 字段
 4. 使用相对方向（画面左侧/右侧），禁止使用东南西北
 5. 角色位置必须与镜头描述一致！
-6. 景深根据 shot_type（全景/中景/近景/特写）自动调整
-7. ⚠️ 对话镜头必须使用浅景深（T2.8或更小），并且注明其他人虚化，确保只有说话者脸部清晰
-8. 如果镜头涉及不同场景，灯光和色调要相应调整
-9. 输出要简洁，每个镜头的规则独立完整
-10. ⚠️ JSON安全：所有引号（""''等）在 JSON 字符串值中必须统一替换为「」，严禁出现未转义的英文双引号 "
+6. 如果角色对象中包含 slot，screen_position / posture / facing 应优先参考该位置语义，但 slot 不是绝对硬限制
+7. 当镜头属于移动过程、入口/出口、过渡区域、路径空间、临时位置、空镜、想象空间、梦境、回忆或抽象空间时，可以基于镜头描述自由决定构图与位置，不必强行贴合 slot
+8. slot 若被引用，必须视为一条完整的位置描述，禁止缩写、改写、总结或替换成短词
+9. 景深根据 shot_type（全景/中景/近景/特写）自动调整
+10. ⚠️ 对话镜头必须使用浅景深（T2.8或更小），并且注明其他人虚化，确保只有说话者脸部清晰
+11. 如果镜头涉及不同场景，灯光和色调要相应调整
+12. 输出要简洁，每个镜头的规则独立完整
+13. ⚠️ JSON安全：所有引号（""''等）在 JSON 字符串值中必须统一替换为「」，严禁出现未转义的英文双引号 "
--- a/lib/prompts/novel-promotion/agent_shot_variant_generate.en.txt
+++ b/lib/prompts/novel-promotion/agent_shot_variant_generate.en.txt
@@ -33,4 +33,5 @@ Execution rules:
 1. Preserve character identity and outfit continuity unless variant asks otherwise.
 2. Preserve location continuity.
 3. Change framing/angle/composition according to target shot and camera move.
-4. Keep one-frame output only, no text overlays.
+4. If characters_info or location_asset includes fixed slots / available slots, keep every visible character anchored to the same fixed slot instead of drifting to another area.
+5. Keep one-frame output only, no text overlays.
--- a/lib/prompts/novel-promotion/agent_shot_variant_generate.zh.txt
+++ b/lib/prompts/novel-promotion/agent_shot_variant_generate.zh.txt
@@ -66,7 +66,8 @@
 2. 保持角色外观与参考图一致（服装、发型、体型）
 3. 保持场景氛围与参考图一致（室内布置、光线、色调）
 4. 改变镜头视角/景别/构图以匹配变体要求
-5. 输出图像比例: {aspect_ratio}
+5. 如果角色信息或场景参考中提供了固定站位 / 可站位置，必须保持人物仍然处于同一固定站位，不得随意换边、换前后景或漂移到其他区域
+6. 输出图像比例: {aspect_ratio}

 ======================================
 【风格要求】
--- a/lib/prompts/novel-promotion/agent_storyboard_detail.en.txt
+++ b/lib/prompts/novel-promotion/agent_storyboard_detail.en.txt
@@ -12,6 +12,8 @@ Location info:

 Task:
 For each panel, output a complete panel object with improved cinematic detail.
+If any character already has `slot`, prefer preserving it exactly and use it as a preferred placement anchor.
+Treat `slot` as one full placement phrase from the location context, not as a short token.

 Required fields per panel:
 - panel_number
@@ -30,7 +32,7 @@ Output schema example (field names must be preserved):
  {
    "panel_number": 1,
    "description": "panel description",
-    "characters": [{ "name": "Character", "appearance": "appearance" }],
+    "characters": [{ "name": "Character", "appearance": "appearance", "slot": "the position beneath the throne steps at the center of the hall" }],
    "location": "location name",
    "scene_type": "daily",
    "source_text": "source text excerpt",
@@ -46,5 +48,9 @@ Rules:
 2. Keep source_text semantically aligned with input; do not rewrite story meaning.
 3. video_prompt should be motion-ready and concrete.
 4. Prefer age+gender wording in video_prompt when naming actors in camera directions.
-5. Return JSON array only.
-6. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
+5. Preserve every input `slot` when the character is stably positioned, and reflect it as a preferred anchor in refined description/video_prompt.
+6. `slot` is not an absolute boundary. If the shot is clearly about movement, entry/exit, path traversal, transition space, temporary space, empty space, imagination, dream, memory, or abstract/non-literal space, you may remove `slot` or keep it without forcing a rigid static match.
+7. When no `slot` is used, decide placement freely from source text, action flow, spatial logic, and cinematic staging.
+8. Do not shorten, rewrite, summarize, or replace a provided `slot` phrase with a short token.
+9. Return JSON array only.
+10. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
--- a/lib/prompts/novel-promotion/agent_storyboard_detail.zh.txt
+++ b/lib/prompts/novel-promotion/agent_storyboard_detail.zh.txt
@@ -5,6 +5,7 @@
 - 为每个分镜设计景别、视角、镜头运动
 - 撰写video_prompt（用年龄段+性别替代角色名）
 - ⚠️ 保留输入分镜中的所有原始字段（特别是 source_text，必须原样保留）
+- 如果输入角色包含 slot，应优先原样保留，并让 refined description/video_prompt 优先参考该位置

 【镜头语言库】

@@ -147,7 +148,7 @@
    "camera_move": "固定",
    "description": "角色A站在桌前，双手撑在桌面上，表情严肃地看着对面的角色B",
    "video_prompt": "年轻男子站在桌前，双手撑在桌面上，表情严肃，正在说话，镜头固定拍摄",
-    "characters": [{"name": "角色A", "appearance": "初始形象"}],
+    "characters": [{"name": "角色A", "appearance": "初始形象", "slot": "皇宫正中龙椅前方台阶下的位置"}],
    "location": "办公室",
    "scene_type": "daily",
    "source_text": "角色A对角色B说：你好"
@@ -179,5 +180,9 @@
 8. 根据输入的分镜数量动态处理
 9. panel_number、characters、location、scene_type保持不变
 10. description可以适当优化，但不要改变核心内容
-11. ⚠️ 必须保留输入分镜中的 source_text 字段，原样输出到结果中，不得遗漏或修改
-12. ⚠️ JSON安全：所有引号（""''等）在 JSON 字符串值中必须统一替换为「」，严禁出现未转义的英文双引号 "
+11. 如果输入中存在 slot，应优先保留并优先参考该位置，但 slot 不是绝对硬边界
+12. 当镜头明显属于移动过程、入口/出口、过渡区域、路径空间、临时位置、空镜、想象空间、梦境、回忆或抽象空间时，可以删除 slot 或保留 slot 但不严格贴合其静态位置
+13. 若不使用 slot，应根据 source_text、动作过程、空间关系与镜头调度自由决定人物位置，不要为了命中 slot 而破坏叙事逻辑
+14. slot 若被保留，必须原样保留为完整位置描述，禁止缩写、改写、总结或替换成短词
+15. ⚠️ 必须保留输入分镜中的 source_text 字段，原样输出到结果中，不得遗漏或修改
+16. ⚠️ JSON安全：所有引号（""''等）在 JSON 字符串值中必须统一替换为「」，严禁出现未转义的英文双引号 "
--- a/lib/prompts/novel-promotion/agent_storyboard_insert.en.txt
+++ b/lib/prompts/novel-promotion/agent_storyboard_insert.en.txt
@@ -23,7 +23,7 @@ Output format (single JSON object only):
 {
  "panel_number": 0,
  "description": "visual description",
-  "characters": [{ "name": "Character Name", "appearance": "appearance name" }],
+  "characters": [{ "name": "Character Name", "appearance": "appearance name", "slot": "the position beneath the throne steps at the center of the hall" }],
  "location": "location name",
  "scene_type": "daily",
  "source_text": "source text or transition shot",
@@ -37,5 +37,8 @@ Rules:
 1. Return one object only (not array).
 2. Keep narrative and spatial continuity between previous and next panel.
 3. Use valid character and location names from provided context.
-4. JSON only, no markdown.
-5. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
+4. If location details include available slots, treat them as preferred anchors. Reuse a provided `slot` for characters who are stably positioned in the scene.
+5. You may omit `slot` when the inserted panel is mainly about movement, entry/exit, path traversal, transition space, temporary space, empty space, imagination, dream, memory, or abstract/non-literal space.
+6. If `slot` is used, it must copy one full placement phrase from the available slots list verbatim. Do not shorten or rename it.
+7. JSON only, no markdown.
+8. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
--- a/lib/prompts/novel-promotion/agent_storyboard_insert.zh.txt
+++ b/lib/prompts/novel-promotion/agent_storyboard_insert.zh.txt
@@ -54,7 +54,7 @@
 |------|------|------|
 | panel_number | number | 固定填 0（由系统重新编号） |
 | description | string | 画面描述：包含角色动作、位置、表情。禁止身份称呼（如"母亲"），使用具体角色名。禁止主观情绪词（如"显得尴尬"），只描述可视化动作。 |
-| characters | array | 出现的角色列表，格式：`[{"name": "角色名", "appearance": "形象名"}]`。角色名必须与角色信息中的名字完全一致。形象名从角色信息的形象列表中选择。 |
+| characters | array | 出现的角色列表，格式：`[{"name": "角色名", "appearance": "形象名", "slot": "场景位置描述"}]`。角色名必须与角色信息中的名字完全一致。形象名从角色信息的形象列表中选择。如果场景信息中提供了可站位置，应优先为稳定停留的角色选择 slot，并直接复用可站位置列表中的完整位置描述。动态移动、过渡区域、入口出口、空镜、想象空间等情况可以不使用 slot。 |
 | location | string | 场景名称，必须与场景信息中的名字完全一致 |
 | scene_type | string | 场景类型，枚举值：`daily`（日常）/ `emotion`（情感）/ `action`（动作）/ `epic`（史诗）/ `suspense`（悬疑） |
 | source_text | string | 对应的原文片段。可以基于前后镜头的 source_text 推断，或填写"过渡镜头" |
@@ -72,6 +72,8 @@
 ❌ location 使用不存在的场景名 → ✅ 必须与场景信息完全一致
 ❌ 特写镜头使用非固定的镜头运动 → ✅ 特写必须用"固定"
 ❌ video_prompt 中使用角色名 → ✅ 必须用年龄段+性别
+❌ 稳定停留位置明明适合使用已有 slot，却完全无视场景锚点 → ✅ 优先复用场景可站位置中的 slot
+❌ 把 slot 改写成短词、代号、缩写 → ✅ 若使用 slot，必须直接复制可站位置列表中的完整位置描述

 ======================================
 【输出格式】
@@ -83,7 +85,7 @@
 {
  "panel_number": 0,
  "description": "...",
-  "characters": [{"name": "...", "appearance": "..."}],
+  "characters": [{"name": "...", "appearance": "...", "slot": "皇宫正中龙椅前方台阶下的位置"}],
  "location": "...",
  "scene_type": "...",
  "source_text": "...",
@@ -91,3 +93,8 @@
  "camera_move": "...",
  "video_prompt": "..."
 }
+
+补充原则：
+- slot 是优先锚点，不是绝对硬边界
+- 当新镜头主要表现角色走动、进入/离开、穿过空间、临时停留、空白空间或想象空间时，可以不使用 slot
+- 若不使用 slot，应根据前后镜头、原文空间关系和过渡逻辑自由决定人物位置
--- a/lib/prompts/novel-promotion/agent_storyboard_plan.en.txt
+++ b/lib/prompts/novel-promotion/agent_storyboard_plan.en.txt
@@ -34,7 +34,7 @@ Output format (JSON array only):
    "panel_number": 1,
    "description": "visual action description",
    "characters": [
-      { "name": "Character Name", "appearance": "appearance name" }
+      { "name": "Character Name", "appearance": "appearance name", "slot": "the position beneath the throne steps at the center of the hall" }
    ],
    "location": "location name",
    "scene_type": "daily",
@@ -52,5 +52,10 @@ Planning rules:
 3. Keep panel transitions smooth and logically continuous.
 4. Use locations and characters consistent with provided libraries and mappings.
 5. Prefer concrete, visible actions over abstract wording.
-6. Return strict JSON only.
-7. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
+6. If location context contains available slots, treat them as preferred placement anchors rather than mandatory hard boundaries.
+7. Prefer assigning a `slot` by copying one full placement phrase from `available_slots` verbatim when a character is stably positioned in the scene.
+8. You may omit `slot` when the shot is mainly about movement, entry/exit, path traversal, transition space, temporary space, empty space, imagination, dream, memory, or abstract/non-literal space.
+9. When no `slot` is used, decide placement freely from source text, action flow, and cinematic staging instead of forcing one of the provided slots.
+10. Do not shorten, rewrite, summarize, or replace a provided slot phrase with a short token such as `slot_1`, `left`, or `throne_front`.
+11. Return strict JSON only.
+12. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
--- a/lib/prompts/novel-promotion/agent_storyboard_plan.zh.txt
+++ b/lib/prompts/novel-promotion/agent_storyboard_plan.zh.txt
@@ -24,9 +24,9 @@
   - 质量优先：确保每个镜头都有意义

 2. 每个分镜必须包含：
-   - panel_number: 分镜序号（1, 2, 3...）
-   - description: 画面描述（人物动作、场景元素、构图要点）
-   - characters: [{name: "角色名", appearance: "形象名"}]
+  - panel_number: 分镜序号（1, 2, 3...）
+  - description: 画面描述（人物动作、场景元素、构图要点）
+   - characters: [{name: "角色名", appearance: "形象名", slot: "场景固定位置描述"}]
   - location: 场景名称（从资产库选择）
   - scene_type: daily/emotion/action/epic/suspense
   - source_text: 对应原文片段 ⚠️ 必填，不得为空
@@ -317,10 +317,13 @@ Clip信息：
 4. 只返回JSON数组，不得有其他文字
 5. ⚠️ source_text 必填，不得为空或null
 6. 空间关系必须清晰（朝向、阻挡、位置）
-7. 镜头连续性：前后镜头要有动作承接
-8. 禁止身份称呼：必须使用资产库中的具体名字
-9. 禁止主观情绪词：只描述可视化动作和状态
-10. 禁止长句单镜头：包含逗号分隔多个动作/对话的长句必须拆分
-11. 对话必须拆分：每段对话至少 2 个镜头（说话者 + 听者反应）
-12. ⚠️ 镜头合理性：只描述当前镜头**实际能拍摄到**的角色，特写/反打等拍不到的可省略
-13. ⚠️ JSON安全：原文中的所有引号（""''「」等）在 JSON 字符串值中必须统一替换为「」，严禁出现未转义的英文双引号 "
+7. 如果场景描述中提供了可站位置列表，应将其视为优先位置锚点，而不是强制覆盖所有可能位置的硬边界
+8. 当角色在场景中稳定停留时，优先从可站位置列表中选择 slot；slot 字段的值必须直接复制完整位置描述，禁止缩写、改写、总结或替换成短词
+9. 当镜头主要表现角色移动、进入/离开、穿过空间、过渡区域、路径空间、临时空间、空白空间、想象/梦境/回忆/抽象空间时，可以不使用 slot，并根据原文和镜头逻辑自由决定位置
+10. 镜头连续性：前后镜头要有动作承接
+11. 禁止身份称呼：必须使用资产库中的具体名字
+12. 禁止主观情绪词：只描述可视化动作和状态
+13. 禁止长句单镜头：包含逗号分隔多个动作/对话的长句必须拆分
+14. 对话必须拆分：每段对话至少 2 个镜头（说话者 + 听者反应）
+15. ⚠️ 镜头合理性：只描述当前镜头**实际能拍摄到**的角色，特写/反打等拍不到的可省略
+16. ⚠️ JSON安全：原文中的所有引号（""''「」等）在 JSON 字符串值中必须统一替换为「」，严禁出现未转义的英文双引号 "
--- a/lib/prompts/novel-promotion/location_create.en.txt
+++ b/lib/prompts/novel-promotion/location_create.en.txt
@@ -7,13 +7,23 @@ User request:
 Rules:
 1. Output in English only.
 2. Start with scene name in this format: "[Scene Name] ..."
-3. Describe a wide, clear environment with spatial layout and key objects.
-4. Mention lighting direction and atmosphere.
-5. No protagonist actions or dialogue.
-6. If crowd is implied by context, use generic crowd terms only (guests, pedestrians, audience).
+3. Describe a wide, complete environment with controllable spatial layout, key structures, and visible depth.
+4. Make foreground, midground, and background explicit.
+5. Define at least 3 clear anchor objects or anchor areas and make the nearby open space visible.
+6. If the user input is generic, such as 「classroom」 or 「office」, proactively make it specific enough for stable image layout instead of staying generic.
+7. Mention lighting direction and atmosphere.
+8. No protagonist actions or dialogue.
+9. If crowd is implied by context, use generic crowd terms only (guests, pedestrians, audience).
+10. Also generate 2-6 fixed `available_slots` as complete descriptive placement phrases tied to concrete scene anchors.
+11. Do not mention posture, action, or emotion in `available_slots`. Describe position only.
+12. Every anchor mentioned in `available_slots` must appear clearly in the scene prompt.

 Output format:
 Return JSON only. ⚠️ JSON SAFETY: All quotation marks MUST be converted to corner brackets「」in JSON string values:
 {
-  "prompt": "[Scene Name] environment description"
+  "prompt": "[Scene Name] environment description",
+  "available_slots": [
+    "the position beside the left edge of the dining table",
+    "the open space just inside the doorway against the wall"
+  ]
 }
--- a/lib/prompts/novel-promotion/location_create.zh.txt
+++ b/lib/prompts/novel-promotion/location_create.zh.txt
@@ -2,7 +2,7 @@

 【场景生成要求（用于出图，中文描述）】

-1. 生成1条中文环境描述（60-120字），像真实摄影场景一样描述
+1. 生成1条中文环境描述（80-140字），像真实摄影场景一样描述，必须足够具体到可以稳定控制画面布局

 2. **开头必须明确写明场景名称**：
   - 描述开头必须以"【场景名称】"的形式标注空间属性
@@ -14,6 +14,10 @@
   - 材质要具体（深棕色实木地板、青灰色石砖墙、做旧铁艺栏杆）
   - 物品要有使用痕迹和生活气息（桌上散落的书籍、墙角堆放的杂物、窗台晒干的植物）
   - 光线要写清楚来源和效果（午后阳光斜照进来在地板上拉出长影、暖黄色壁灯打在墙面上）
+   - 必须写出完整空间结构，不要只写泛场景名词
+   - 必须写出前景/中景/背景或近处/中部/远处层次
+   - 必须写出至少3个清晰可见的关键锚点及其周边空位
+   - 如果用户输入很泛（如「学校教室」「办公室」），你必须主动把它具体化为可控构图，而不是停留在泛描述

 4. 禁止：不写主角人物具体动作、不写画风、不写"温馨""优雅"等抽象词

@@ -22,9 +26,20 @@
   - 人群描述示例："大厅中宾客三两成群"、"街道上行人往来"、"座位上零散坐着几位观众"
   - 如果是私密空间或用户明确要求空镜，则不添加人群

+6. 额外生成 2-6 个该场景的固定可站位置：
+   - 每个位置必须是一条完整的位置描述短语，而不是短词
+   - 每个位置必须依附于明确的场景锚物或区域
+   - 位置描述中禁止写人物姿态、动作、情绪，只写空间位置
+   - 这些位置里提到的锚点必须在场景描述中真实出现
+   - 示例：饭桌左侧靠桌边的位置、教室后排靠窗那组课桌外侧的位置、皇宫正中龙椅前方台阶下的位置
+
 以下是用户的生成指令:{user_input}

 只返回以下json格式，禁止返回一切除json以外的多余内容。⚠️ 所有引号（""''等）在 JSON 字符串值中必须替换为「」，严禁出现未转义的英文双引号 "。
 {
-  "prompt":"「场景名称」场景描述内容"
+  "prompt":"「场景名称」场景描述内容",
+  "available_slots":[
+    "饭桌左侧靠桌边的位置",
+    "门口内侧靠墙的位置"
+  ]
 }
--- a/lib/prompts/novel-promotion/location_description_update.en.txt
+++ b/lib/prompts/novel-promotion/location_description_update.en.txt
@@ -18,9 +18,18 @@ Rules:
 2. Return one complete updated description in English.
 3. Keep scene name at the beginning: "[{location_name}] ..."
 4. No protagonist actions or story narration.
+5. Keep the scene spatially specific, with visible structure, depth, and stable anchors.
+6. Also regenerate 2-6 `available_slots` that match the updated scene.
+7. Each `available_slots` item must be one complete descriptive placement phrase, not a short token and not an object.
+8. Do not mention posture, action, or emotion in `available_slots`.
+9. Every anchor mentioned in `available_slots` must also appear clearly in the updated scene description.

 Output format:
 Return JSON only. ⚠️ JSON SAFETY: All quotation marks MUST be converted to corner brackets「」in JSON string values:
 {
-  "prompt": "updated location description"
+  "prompt": "updated location description",
+  "available_slots": [
+    "the outer-side position beside the rear window desks",
+    "the open floor directly below the center of the blackboard"
+  ]
 }
--- a/lib/prompts/novel-promotion/location_description_update.zh.txt
+++ b/lib/prompts/novel-promotion/location_description_update.zh.txt
@@ -27,10 +27,20 @@
 6. 保留未被修改的原有特征
 7. 遵循以下描述规范：
   - 只描述场景本身，禁止描述人物
-   - 使用中文输出，长度 50-100 字
+   - 使用中文输出，长度 80-140 字
+   - 必须让空间结构、关键锚点、前后层次具体可见，不能退化成泛场景描述
+   - 若用户修改后引入新的关键锚点或删掉旧锚点，描述必须同步更新
+8. 同时重新生成 2-6 个固定可站位置，且必须与更新后的场景描述一致
+9. 每个可站位置必须是一条完整的位置描述短语，不是短词，不是对象
+10. 可站位置中禁止写人物姿态、动作、情绪，只写位置
+11. 可站位置里提到的关键锚点必须在更新后的场景描述中明确出现

 【输出格式】
 只返回JSON格式，禁止返回任何其他内容。⚠️ 所有引号（""''等）在 JSON 字符串值中必须替换为「」，严禁出现未转义的英文双引号 "：
 {
-  "prompt": "「场景名」更新后的完整场景描述"
+  "prompt": "「场景名」更新后的完整场景描述",
+  "available_slots":[
+    "教室后排靠窗那组课桌外侧的位置",
+    "讲台前方黑板正下方的位置"
+  ]
 }
--- a/lib/prompts/novel-promotion/location_modify.en.txt
+++ b/lib/prompts/novel-promotion/location_modify.en.txt
@@ -16,9 +16,19 @@ Rules:
 3. Output in English only.
 4. Start with scene name: "[{location_name}] ..."
 5. No protagonist actions, dialogue, or narrative plot.
+6. Make the scene spatially specific enough for controllable image generation, with visible structure, depth, and stable anchors.
+7. Keep at least 3 clear anchor objects or anchor areas visible in the modified scene.
+8. Regenerate 2-6 `available_slots` that stay consistent with the modified scene.
+9. Each `available_slots` item must be one complete descriptive placement phrase, not a short token and not an object.
+10. Do not mention posture, action, or emotion in `available_slots`.
+11. Every anchor mentioned in `available_slots` must also appear in the modified scene description.

 Output format:
 Return JSON only. ⚠️ JSON SAFETY: All quotation marks MUST be converted to corner brackets「」in JSON string values:
 {
-  "prompt": "modified location description"
+  "prompt": "modified location description",
+  "available_slots": [
+    "the position beneath the throne steps at the center of the hall",
+    "the open space between the left column and the long table"
+  ]
 }
--- a/lib/prompts/novel-promotion/location_modify.zh.txt
+++ b/lib/prompts/novel-promotion/location_modify.zh.txt
@@ -45,6 +45,9 @@
     * ✅ 好："木质书桌"/"灰色布艺沙发"/"白色窗帘"
     * ❌ 差："优雅的家具"/"舒适的环境"/"温馨的氛围"
   - 描述固定元素，不写主角人物具体动作、情绪
+   - 必须让空间结构足够具体，不能只剩泛场景标签
+   - 必须让前景/中景/背景或近处/中部/远处关系清楚
+   - 必须保留至少 3 个清晰可见、可供后续落位的锚点或区域
   - 【人群处理规则】如果用户要求添加人群，或场景本身暗示有人群（如宴会、集市等）：
     * 可以加入模糊人群描述：\"人群\"、\"宾客\"、\"路人\"等
     * 示例：\"大厅远处三两宾客交谈\"、\"街角有行人匆匆走过\"
@@ -54,11 +57,22 @@

 你的目标是根据用户的修改指令，在原有场景描述的基础上进行修改

+同时你必须重新生成该场景的固定可站位置：
+- 输出 2-6 个站位
+- 每个站位必须是一条完整的位置描述短语，不是短词
+- 每个站位必须依附于明确场景锚物或区域
+- 站位中禁止写人物姿态、动作、情绪，只写位置
+- 新站位必须与修改后的场景描述一致，且其中提到的锚点必须在场景描述中出现
+
 当前场景描述：{location_input}

 用户的修改指令：{user_input}

 发送json格式给我，只返回以下json格式，禁止返回一切除json以外的多余内容，注释，文字等等，只返回无任何markdown标识符的纯净json格式。⚠️ 所有引号（""''等）在 JSON 字符串值中必须统一替换为「」，严禁出现未转义的英文双引号 "。json格式如下
 {
-  "prompt":"「场景名」xxxxx"
+  "prompt":"「场景名」xxxxx",
+  "available_slots":[
+    "皇宫正中龙椅前方台阶下的位置",
+    "左侧立柱与长案之间的空位"
+  ]
 }
--- a/lib/prompts/novel-promotion/location_regenerate.en.txt
+++ b/lib/prompts/novel-promotion/location_regenerate.en.txt
@@ -12,7 +12,11 @@ Requirements:
 2. Keep the scene name prefix in each line: "[{location_name}] ..."
 3. Output in English only.
 4. Keep environment-only description (no protagonist actions).
-5. Keep each variant concise and image-generation friendly.
+5. Keep each variant specific enough for controllable image generation, with visible structure, depth, and stable anchors.
+6. Also generate 2-6 shared `available_slots` for this location.
+7. Each `available_slots` item must be one complete descriptive placement phrase, not a short token and not an object.
+8. Do not mention posture, action, or emotion in `available_slots`.
+9. Every anchor mentioned in `available_slots` must remain valid across all three regenerated descriptions.

 Output format:
 Return JSON only. ⚠️ JSON SAFETY: All quotation marks MUST be converted to corner brackets「」in JSON string values:
@@ -21,5 +25,9 @@ Return JSON only. ⚠️ JSON SAFETY: All quotation marks MUST be converted to c
    "[{location_name}] variant 1",
    "[{location_name}] variant 2",
    "[{location_name}] variant 3"
+  ],
+  "available_slots": [
+    "the position beneath the throne steps at the center of the hall",
+    "the position against the inner wall beside the rear doorway"
  ]
 }
--- a/lib/prompts/novel-promotion/location_regenerate.zh.txt
+++ b/lib/prompts/novel-promotion/location_regenerate.zh.txt
@@ -20,6 +20,8 @@
   - 材质细节：地面、墙面、物体的材质质感
   - 环境元素：植物、天气、装饰物等
   - 独特标识：该场景的标志性元素或特殊物件
+   - 至少 3 个可供后续固定人物位置的稳定锚点或区域
+   - 每个锚点周边可落位的空白区域

 5. 描述规范：
   - 禁止写主角人物具体动作、剧情
@@ -31,11 +33,23 @@
   - 【年代一致性】根据场景特征判断年代，建筑、装饰、物品必须符合该年代特征
   - 【时间一致性】如场景名包含"白天/黑夜/黄昏"等，描述中的光影必须匹配

+6. 额外输出 2-6 个固定可站位置：
+   - 与该场景的共通构图一致
+   - 每个站位必须是一条完整的位置描述短语，不是短词，不是对象
+   - 依附于明确场景锚物或区域
+   - 禁止抽象站位
+   - 禁止写人物姿态、动作、情绪
+   - 站位中提到的锚点必须在三条 descriptions 中都成立
+
 【输出格式】只返回以下 JSON，不要任何其他内容。⚠️ 所有引号（""''等）在 JSON 字符串值中必须替换为「」，严禁出现未转义的英文双引号 "。
 {
  "descriptions": [
    "「场景名」新描述1（80-150字）",
    "「场景名」新描述2（80-150字）",
    "「场景名」新描述3（80-150字）"
+  ],
+  "available_slots": [
+    "皇宫正中龙椅前方台阶下的位置",
+    "右后方殿门内侧靠墙的位置"
  ]
 }
--- a/lib/prompts/novel-promotion/select_location.en.txt
+++ b/lib/prompts/novel-promotion/select_location.en.txt
@@ -17,9 +17,17 @@ For each selected location, generate 3 wide-angle environment descriptions.
 Each description should:
 - start with location name in brackets: "[Location Name] ..."
 - describe spatial layout, depth layers, major objects, and lighting direction
+- define at least 3 stable anchor objects or anchor areas that can support later character placement
+- make the usable open space around those anchors visually clear
 - remain environment-only (no named protagonist actions)
 - use concise, production-ready English

+If the source location is generic, such as 「classroom」, 「office」, or 「living room」, you must proactively make it specific enough for controllable image generation:
+- define the visible main structure
+- define foreground, midground, and background
+- define at least 3 stable anchor areas and nearby open space
+- avoid vague noun piles
+
 Output format (JSON only):
 {
  "locations": [
@@ -28,6 +36,11 @@ Output format (JSON only):
      "summary": "short usage summary",
      "has_crowd": false,
      "crowd_description": "",
+      "available_slots": [
+        "the position beneath the throne steps at the center of the palace hall",
+        "the open space between the left column and the long table",
+        "the position against the inner wall beside the rear doorway"
+      ],
      "descriptions": [
        "[location_name] description 1",
        "[location_name] description 2",
@@ -40,5 +53,10 @@ Output format (JSON only):
 Strict constraints:
 1. JSON only.
 2. If no valid location exists, return: {"locations":[]}.
-3. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
-  
+3. Each location must include 2-6 `available_slots`.
+4. Each `available_slots` item must be a complete descriptive placement phrase, not a short token and not an object.
+5. Slot descriptions must be concrete and reusable, tied to visible anchors such as the outer side of a desk row, the open floor before the blackboard, the inner side of a doorway, the space beside a window wall.
+6. Do not mention character posture, action, or emotion inside `available_slots`. Describe position only.
+7. Every anchor mentioned in `available_slots` must also appear clearly in the location descriptions.
+8. ⚠️ JSON SAFETY: All quotation marks in text (""''「」 etc.) MUST be converted to corner brackets「」in JSON string values. NEVER use raw ASCII double quotes " inside string values.
+  
--- a/lib/prompts/novel-promotion/select_location.zh.txt
+++ b/lib/prompts/novel-promotion/select_location.zh.txt
@@ -72,9 +72,15 @@
   - 使用明确的位置词：左侧/右侧/中央/角落/靠窗/远处
   - 描述物体之间的空间关系和前后层次
   - 5-8件物体，每件都有位置说明
+   - 至少 2-3 个后续可作为人物落位锚点的关键物体或区域必须被明确写出，如桌边、门内侧、窗下墙边、讲台前、龙椅前台阶

   **光线方向**：光从哪个方向照入，照亮哪些区域

+   **可落位空间**（必须体现）：
+   - 必须说明哪些区域留有可供人物站立或出现的空白空间
+   - 这些空白区域必须与关键锚点相邻，便于后续固定人物位置
+   - 禁止把所有锚点都塞满家具或遮挡物，导致后续无法落人
+
 5. 描述规范：
   - 强调位置关系词：前方、远处、左侧、角落、靠近、深处
   - 长度 100-150 字
@@ -108,6 +114,12 @@

 8.如无特殊要求，使用用户输入的语言来进行场景生成，例如输入英文输出偏西方场景，中文则输出偏中国场景，但是原则要按照文字剧本里实际发生的地点为准，

+9. 如果原文或用户输入过于泛化（如「学校教室」「办公室」「客厅」），你必须主动将其具体化为可控画面的完整空间：
+   - 明确主视角下能看到的关键结构
+   - 明确前景/中景/背景
+   - 明确至少 3 个稳定锚点及其周边空位
+   - 禁止只输出泛泛的场景名词堆砌
+
 【输出规范（只允许以下 JSON 结构；字段名中文；不得输出任何多余文字）】
 {
  "locations": [
@@ -116,6 +128,11 @@
      "summary": "场景简要说明（用途/人物关联，如：张三居住的主卧室、公司高层会议室等）",
      "has_crowd": true/false,
      "crowd_description": "人群类型描述（仅当has_crowd为true时填写，如：宴会宾客、集市人群、学生们等）",
+      "available_slots": [
+        "皇宫正中龙椅前方台阶下的位置",
+        "左侧立柱与长案之间的空位",
+        "右后方殿门内侧靠墙的位置"
+      ],
      "descriptions": [
        "「场景名」场景环境描述1（如has_crowd为true则包含人群元素）",
        "「场景名」场景环境描述2",
@@ -127,6 +144,11 @@

 【严格性】
 - 若无符合条件的场景，locations数组返回 []。
+- 每个场景必须生成 2-6 个 available_slots，且每个站位都必须具体、可复用、与场景内明确锚物相关。
+- 每个 available_slots 元素必须是一条完整的位置描述短语，不是短词，不是结构化对象。
+- 站位描述必须像「皇宫正中龙椅前方台阶下的位置」「教室后排靠窗那组课桌外侧的位置」这样，直接说明锚物、方位和具体区域。
+- 禁止抽象站位，如「左边」「中间」「角落」；禁止写人物姿态、动作、情绪；只描述位置本身。
+- available_slots 中提到的所有关键锚点，必须在 descriptions 中清楚出现，否则该站位无效，不能输出。
 - 只返回上述 JSON；不得输出markdown代码块标记、如```json注释或解释；不得添加未定义字段。
 - 每条描述必须遵守长度限制（100-150字）；发现超长请自行截断。
 - 禁止在 JSON 字符串值中出现英文双引号 "。原文中的所有引号（""''等）必须统一替换为「」。如字符串内确实需要英文双引号，必须转义为 \"
--- a/lib/prompts/novel-promotion/single_panel_image.en.txt
+++ b/lib/prompts/novel-promotion/single_panel_image.en.txt
@@ -21,7 +21,10 @@ Style requirement:

 Execution rules:
 1. Respect panel composition, character placement, and action logic.
-2. Use reference images for style/identity consistency only.
-3. Repaint the background according to shot type and angle.
-4. If storyboard conflicts with source text, keep narrative logic from source text.
-5. Keep final visual style consistent with provided references.
+2. If storyboard data contains `slot`, treat it as a preferred placement anchor for consistency, not as an absolute restriction.
+3. If location data contains `available_slots`, treat them as typical anchor references rather than a complete map of all valid positions.
+4. When the panel description, source text, action flow, or scene nature implies movement, entry/exit, path traversal, transition space, temporary space, empty space, or abstract/non-literal space, do not force the character to remain inside an existing slot.
+5. Use reference images for style/identity consistency only.
+6. Repaint the background according to shot type and angle.
+7. If storyboard conflicts with source text, keep narrative logic from source text.
+8. Keep final visual style consistent with provided references.
--- a/lib/prompts/novel-promotion/single_panel_image.zh.txt
+++ b/lib/prompts/novel-promotion/single_panel_image.zh.txt
@@ -50,6 +50,10 @@
 - 严格按照分镜要求绘制画面
 - 禁止添加、删除或重排任何镜头
 - 镜头必须与输入完全匹配
+- 如果分镜数据中包含角色 slot，可将其视为优先位置参考，用于保持人物落位一致性，但不是绝对硬限制
+- 如果场景数据中包含 available_slots，应将其理解为典型位置参考，而不是完整空间边界
+- 当镜头描述、原文空间关系、动作过程或场景性质表明人物正在移动、处于过渡区域、入口出口、路径空间、临时空间、空白空间或想象/梦境/回忆/抽象空间时，不要强行把人物锁死在现有 slot 中
+- 最终以镜头叙事正确、空间关系自然、人物位置合理为最高原则

 【分镜数据】
 {storyboard_text_json_input}