feat: 改用官网拼接延长效果更佳

app/api/v1/public_api/video.py

@@ -139,6 +139,13 @@ async def _new_session(
     parent_post_id: Optional[str],
     source_image_url: Optional[str],
     reasoning_effort: Optional[str],
+    # 视频延长相关
+    is_video_extension: bool = False,
+    extend_post_id: Optional[str] = None,
+    video_extension_start_time: Optional[float] = None,
+    original_post_id: Optional[str] = None,
+    file_attachment_id: Optional[str] = None,
+    stitch_with_extend: bool = True,
 ) -> str:
     task_id = uuid.uuid4().hex
     now = time.time()
@@ -154,6 +161,12 @@ async def _new_session(
             "parent_post_id": parent_post_id,
             "source_image_url": source_image_url,
             "reasoning_effort": reasoning_effort,
+            "is_video_extension": is_video_extension,
+            "extend_post_id": extend_post_id,
+            "video_extension_start_time": video_extension_start_time,
+            "original_post_id": original_post_id,
+            "file_attachment_id": file_attachment_id,
+            "stitch_with_extend": stitch_with_extend,
             "created_at": now,
         }
     return task_id
@@ -234,6 +247,13 @@ class VideoStartRequest(BaseModel):
     source_image_url: Optional[str] = None
     reasoning_effort: Optional[str] = None
     edit_context: Optional[Dict[str, Any]] = None
+    # 视频延长相关字段
+    is_video_extension: Optional[bool] = False
+    extend_post_id: Optional[str] = None
+    video_extension_start_time: Optional[float] = None
+    original_post_id: Optional[str] = None
+    file_attachment_id: Optional[str] = None
+    stitch_with_extend: Optional[bool] = True
 
 
 @router.post("/video/start", dependencies=[Depends(verify_public_key)])
@@ -281,15 +301,47 @@ async def public_video_start(data: VideoStartRequest):
     elif source_image_url:
         _validate_image_url(source_image_url)
 
-    if parent_post_id and image_url:
-        raise HTTPException(
-            status_code=400, detail="image_url and parent_post_id cannot be used together"
-        )
-    if not prompt and not image_url and not parent_post_id:
-        raise HTTPException(
-            status_code=400,
-            detail="Prompt cannot be empty when no image_url/parent_post_id is provided",
+    # 视频延长参数解析
+    is_video_extension = bool(data.is_video_extension)
+    extend_post_id = _validate_parent_post_id(data.extend_post_id or "")
+    video_extension_start_time = data.video_extension_start_time
+    original_post_id = _validate_parent_post_id(data.original_post_id or "")
+    file_attachment_id = _validate_parent_post_id(data.file_attachment_id or "")
+    stitch_with_extend = bool(data.stitch_with_extend if data.stitch_with_extend is not None else True)
+
+    if is_video_extension:
+        # 视频延长模式校验
+        if not extend_post_id:
+            raise HTTPException(
+                status_code=400,
+                detail="extend_post_id is required for video extension",
+            )
+        if video_extension_start_time is None or video_extension_start_time < 0:
+            raise HTTPException(
+                status_code=400,
+                detail="video_extension_start_time must be a non-negative number",
+            )
+        
+        # 官方服务端对延长视频容易触发风控，在此强制固定并发为1
+        concurrent = 1
+
+        logger.info(
+            "Public video extension request: "
+            f"extend_post_id={extend_post_id}, "
+            f"start_time={video_extension_start_time}, "
+            f"original_post_id={original_post_id}, "
+            f"file_attachment_id={file_attachment_id}"
         )
+    else:
+        if parent_post_id and image_url:
+            raise HTTPException(
+                status_code=400, detail="image_url and parent_post_id cannot be used together"
+            )
+        if not prompt and not image_url and not parent_post_id:
+            raise HTTPException(
+                status_code=400,
+                detail="Prompt cannot be empty when no image_url/parent_post_id is provided",
+            )
 
     reasoning_effort = (data.reasoning_effort or "").strip() or None
     if reasoning_effort:
@@ -326,6 +378,12 @@ async def public_video_start(data: VideoStartRequest):
             parent_post_id,
             source_image_url,
             reasoning_effort,
+            is_video_extension=is_video_extension,
+            extend_post_id=extend_post_id,
+            video_extension_start_time=video_extension_start_time,
+            original_post_id=original_post_id,
+            file_attachment_id=file_attachment_id,
+            stitch_with_extend=stitch_with_extend,
         )
         task_ids.append(task_id)
 
@@ -335,6 +393,8 @@ async def public_video_start(data: VideoStartRequest):
         "concurrent": concurrent,
         "aspect_ratio": aspect_ratio,
         "parent_post_id": parent_post_id,
+        "extend_post_id": extend_post_id,
+        "file_attachment_id": file_attachment_id or "",
     }
 
 
@@ -401,6 +461,14 @@ async def public_video_sse(request: Request, task_id: str = Query("")):
             else:
                 messages = [{"role": "user", "content": prompt}]
 
+            # 从 session 取得视频延长参数
+            is_video_extension = bool(session.get("is_video_extension"))
+            extend_post_id = str(session.get("extend_post_id") or "").strip() or None
+            video_extension_start_time = session.get("video_extension_start_time")
+            original_post_id = str(session.get("original_post_id") or "").strip() or None
+            file_attachment_id = str(session.get("file_attachment_id") or "").strip() or None
+            stitch_with_extend = bool(session.get("stitch_with_extend", True))
+
             stream = await VideoService.completions(
                 model_id,
                 messages,
@@ -411,12 +479,21 @@ async def public_video_sse(request: Request, task_id: str = Query("")):
                 resolution=resolution_name,
                 preset=preset,
                 parent_post_id=parent_post_id or None,
+                extend_post_id=extend_post_id if is_video_extension else None,
+                video_extension_start_time=video_extension_start_time if is_video_extension else None,
+                original_post_id=original_post_id if is_video_extension else None,
+                file_attachment_id=file_attachment_id if is_video_extension else None,
+                stitch_with_extend=stitch_with_extend,
                 source_image_url=source_image_url,
                 preferred_token=preferred_token,
             )
 
             async for chunk in stream:
                 if await request.is_disconnected():
+                    logger.info(f"Public video client disconnected: {task_id}")
+                    break
+                if task_id not in _VIDEO_SESSIONS:
+                    logger.info(f"Public video task stopped by user: {task_id}")
                     break
                 yield chunk
         except Exception as e:

app/services/grok/services/image.py

@@ -425,13 +425,30 @@ class ImageWSBaseProcessor(BaseProcessor):
     async def _to_output(self, image_id: str, item: Dict) -> str:
         try:
             if self.response_format == "url":
-                return await self._save_blob(
+                res = await self._save_blob(
                     image_id,
                     item.get("blob", ""),
                     item.get("is_final", False),
                     ext=item.get("ext"),
                 )
-            return self._strip_base64(item.get("blob", ""))
+                
+                # [NEW] 当图像生成存盘时，保存其 tokenId 用以后续延长关联
+                if res and image_id and getattr(self, "token", None):
+                    from app.services.grok.utils.asset_token_map import AssetTokenMap
+                    token_map = await AssetTokenMap.get_instance()
+                    await token_map.save_mapping(image_id, self.token)
+                    
+                return res
+                
+            res = self._strip_base64(item.get("blob", ""))
+            
+            # [NEW] Base64 响应下同样强绑定
+            if res and image_id and getattr(self, "token", None):
+                from app.services.grok.utils.asset_token_map import AssetTokenMap
+                token_map = await AssetTokenMap.get_instance()
+                await token_map.save_mapping(image_id, self.token)
+                
+            return res
         except Exception as e:
             logger.warning(f"Image output failed: {e}")
             return ""

app/services/grok/services/video.py

@@ -167,6 +167,17 @@ class VideoService:
 
         return True
 
+    @staticmethod
+    def _map_preset_to_mode(preset: str) -> str:
+        """将前端预设名映射为 Grok 官方 mode 参数。"""
+        mapping = {
+            "spicy": "extremely-spicy-or-crazy",
+            "fun": "extremely-crazy",
+            "normal": "normal",
+        }
+        # 如果预设没发送，默认 --mode=extremely-spicy-or-crazy (即 spicy)
+        return mapping.get(preset, "extremely-spicy-or-crazy")
+
     @staticmethod
     def _build_video_message(
         prompt: str,
@@ -175,18 +186,22 @@ class VideoService:
     ) -> str:
         """构造视频请求 message：
         - 有提示词：统一走 custom，并发送 image_url + prompt + mode
-        - 无提示词：统一走 spicy（忽略 preset）
+        - 无提示词：根据所选 preset 转换 mode
         """
         prompt_text = (prompt or "").strip()
         if not VideoService.is_meaningful_video_prompt(prompt_text):
             prompt_text = ""
+
         image_core = (source_image_url or "").strip()
         if prompt_text:
+            mode_flag = "--mode=custom"
             if image_core:
-                return f"{image_core}  {prompt_text} --mode=custom"
-            return f"{prompt_text} --mode=custom"
+                return f"{image_core}  {prompt_text} {mode_flag}"
+            return f"{prompt_text} {mode_flag}"
 
-        mode_flag = "--mode=extremely-spicy-or-crazy"
+        # 无提示词（或泛化指令）
+        official_mode = VideoService._map_preset_to_mode(preset)
+        mode_flag = f"--mode={official_mode}"
         if image_core:
             return f"{image_core}  {mode_flag}"
         return mode_flag
@@ -269,13 +284,12 @@ class VideoService:
     ) -> AsyncGenerator[bytes, None]:
         """Generate video."""
         token_tag = _token_tag(token)
-        mode = (
-            "custom"
-            if VideoService.is_meaningful_video_prompt(prompt)
-            else "extremely-spicy-or-crazy"
-        )
+        # 确定逻辑上的 mode
+        is_custom = VideoService.is_meaningful_video_prompt(prompt)
+        official_mode = "custom" if is_custom else VideoService._map_preset_to_mode(preset)
+
         logger.info(
-            f"Video generation: token={token_tag}, prompt='{prompt[:50]}...', ratio={aspect_ratio}, length={video_length}s, mode={mode}"
+            f"Video generation: token={token_tag}, prompt='{prompt[:50]}...', ratio={aspect_ratio}, length={video_length}s, mode={official_mode}"
         )
         post_id = await self.create_post(token, prompt)
         message = self._build_video_message(prompt=prompt, preset=preset)
@@ -303,6 +317,7 @@ class VideoService:
                             token,
                             message=message,
                             model="grok-3",
+                            mode=official_mode,
                             tool_overrides={"videoGen": True},
                             model_config_override=model_config_override,
                         )
@@ -359,13 +374,12 @@ class VideoService:
     ) -> AsyncGenerator[bytes, None]:
         """Generate video from image."""
         token_tag = _token_tag(token)
-        mode = (
-            "custom"
-            if VideoService.is_meaningful_video_prompt(prompt)
-            else "extremely-spicy-or-crazy"
-        )
+        # 确定逻辑上的 mode
+        is_custom = VideoService.is_meaningful_video_prompt(prompt)
+        official_mode = "custom" if is_custom else VideoService._map_preset_to_mode(preset)
+
         logger.info(
-            f"Image to video: token={token_tag}, prompt='{prompt[:50]}...', image={image_url[:80]}, mode={mode}"
+            f"Image to video: token={token_tag}, prompt='{prompt[:50]}...', image={image_url[:80]}, mode={official_mode}"
         )
         post_id = await self.create_image_post(token, image_url)
         message = self._build_video_message(
@@ -397,6 +411,7 @@ class VideoService:
                             token,
                             message=message,
                             model="grok-3",
+                            mode=official_mode,
                             tool_overrides={"videoGen": True},
                             model_config_override=model_config_override,
                         )
@@ -454,11 +469,7 @@ class VideoService:
     ) -> AsyncGenerator[bytes, None]:
         """Generate video by existing parent post ID (preferred path)."""
         token_tag = _token_tag(token)
-        mode = (
-            "custom"
-            if VideoService.is_meaningful_video_prompt(prompt)
-            else "extremely-spicy-or-crazy"
-        )
+        is_custom = VideoService.is_meaningful_video_prompt(prompt)
         logger.info(
             f"ParentPost to video: token={token_tag}, prompt='{prompt[:50]}...', parent_post_id={parent_post_id}"
         )
@@ -504,11 +515,15 @@ class VideoService:
         }
         moderated_max_retry = max(1, int(get_config("video.moderated_max_retry", 5)))
 
+        # 确定逻辑上的 mode
+        is_custom = VideoService.is_meaningful_video_prompt(prompt)
+        official_mode = "custom" if is_custom else VideoService._map_preset_to_mode(preset)
+
         logger.info(
             "ParentPost video request prepared: "
             f"token={token_tag}, parent_post_id={parent_post_id}, "
-            f"message_len={len(message)}, has_prompt={bool((prompt or '').strip())}, "
-            f"resolution={resolution}, video_length={video_length}, ratio={aspect_ratio}, mode={mode}"
+            f"message_len={len(message)}, has_prompt={is_custom}, "
+            f"resolution={resolution}, video_length={video_length}, ratio={aspect_ratio}, mode={official_mode}"
         )
 
         async def _stream():
@@ -522,6 +537,7 @@ class VideoService:
                             token,
                             message=message,
                             model="grok-3",
+                            mode=official_mode,
                             tool_overrides={"videoGen": True},
                             model_config_override=model_config_override,
                         )
@@ -567,6 +583,144 @@ class VideoService:
 
         return _stream()
 
+    async def generate_extend_video(
+        self,
+        token: str,
+        prompt: str,
+        extend_post_id: str,
+        video_extension_start_time: float,
+        original_post_id: str = "",
+        file_attachment_id: str = "",
+        aspect_ratio: str = "16:9",
+        video_length: int = 6,
+        resolution: str = "480p",
+        preset: str = "normal",
+        stitch_with_extend: bool = True,
+    ) -> AsyncGenerator[bytes, None]:
+        """通过 Grok 官方视频延长 API 延长视频。"""
+        token_tag = _token_tag(token)
+        # 确定 mode
+        prompt_text = (prompt or "").strip()
+        is_custom = VideoService.is_meaningful_video_prompt(prompt_text)
+        if is_custom:
+            mode = "custom"
+        else:
+            mode = VideoService._map_preset_to_mode(preset)
+            prompt_text = ""
+
+        effective_original = (original_post_id or "").strip() or extend_post_id
+        effective_file_attachment = (file_attachment_id or "").strip() or effective_original
+
+        logger.info(
+            "Video extension request: "
+            f"token={token_tag}, extend_post_id={extend_post_id}, "
+            f"start_time={video_extension_start_time}, original_post_id={effective_original}, "
+            f"prompt='{(prompt_text or '')[:50]}', mode={mode}"
+        )
+
+        # 构造 message
+        if prompt_text:
+            message = f"{prompt_text} --mode={mode}"
+        else:
+            message = f"--mode={mode}"
+
+        # 构造 videoGenModelConfig —— 对齐官网抓包格式
+        video_gen_config = {
+            "isVideoExtension": True,
+            "videoExtensionStartTime": video_extension_start_time,
+            "extendPostId": extend_post_id,
+            "stitchWithExtendPostId": stitch_with_extend,
+            "originalPostId": effective_original,
+            "originalRefType": "ORIGINAL_REF_TYPE_VIDEO_EXTENSION",
+            "mode": mode,
+            "aspectRatio": aspect_ratio,
+            "videoLength": video_length,
+            "resolutionName": resolution,
+            "parentPostId": extend_post_id,
+            "isVideoEdit": False,
+        }
+        if prompt_text:
+            video_gen_config["originalPrompt"] = prompt_text
+
+        model_config_override = {
+            "modelMap": {
+                "videoGenModelConfig": video_gen_config,
+            }
+        }
+
+        # fileAttachments 对齐官网：始终传最初图转视频时的 parentPostId
+        file_attachments = [effective_file_attachment]
+
+        moderated_max_retry = max(1, int(get_config("video.moderated_max_retry", 5)))
+
+        logger.info(
+            "Video extension request prepared: "
+            f"token={token_tag}, extend_post_id={extend_post_id}, "
+            f"file_attachments={file_attachments}, "
+            f"start_time={video_extension_start_time}, mode={mode}, "
+            f"resolution={resolution}, video_length={video_length}, ratio={aspect_ratio}"
+        )
+
+        async def _stream():
+            for attempt in range(1, moderated_max_retry + 1):
+                session = AsyncSession()
+                moderated_hit = False
+                try:
+                    async with _get_video_semaphore():
+                        stream_response = await AppChatReverse.request(
+                            session,
+                            token,
+                            message=message,
+                            model="grok-3",
+                            mode=mode,
+                            file_attachments=file_attachments,
+                            tool_overrides={"videoGen": True},
+                            model_config_override=model_config_override,
+                        )
+                        logger.info(
+                            "Video extension started: "
+                            f"token={token_tag}, extend_post_id={extend_post_id}, "
+                            f"attempt={attempt}/{moderated_max_retry}"
+                        )
+                        async for line in stream_response:
+                            if self._is_moderated_line(line):
+                                moderated_hit = True
+                                logger.warning(
+                                    f"Video extension moderated: token={token_tag}, "
+                                    f"retry {attempt}/{moderated_max_retry}"
+                                )
+                                break
+                            yield line
+
+                    if not moderated_hit:
+                        return
+                    if attempt < moderated_max_retry:
+                        await asyncio.sleep(1.2)
+                        continue
+                    raise UpstreamException(
+                        "Video extension blocked by moderation",
+                        status_code=400,
+                        details={"moderated": True, "attempts": moderated_max_retry},
+                    )
+                except Exception as e:
+                    logger.error(f"Video extension error: {e}")
+                    if isinstance(e, AppException):
+                        raise
+                    msg, code, status = _classify_video_error(e)
+                    raise AppException(
+                        message=msg,
+                        error_type=ErrorType.SERVER.value if status >= 500 else ErrorType.INVALID_REQUEST.value,
+                        code=code,
+                        status_code=status,
+                    )
+                finally:
+                    try:
+                        await session.close()
+                    except Exception:
+                        pass
+
+        return _stream()
+
     @staticmethod
     async def completions(
         model: str,
@@ -578,6 +732,11 @@ class VideoService:
         resolution: str = "480p",
         preset: str = "normal",
         parent_post_id: str | None = None,
+        extend_post_id: str | None = None,
+        video_extension_start_time: float | None = None,
+        original_post_id: str | None = None,
+        file_attachment_id: str | None = None,
+        stitch_with_extend: bool = True,
         source_image_url: str | None = None,
         preferred_token: str | None = None,
     ):
@@ -603,6 +762,19 @@ class VideoService:
         parent_post_id = (parent_post_id or "").strip() or None
         source_image_url = (source_image_url or "").strip()
         preferred_token = (preferred_token or "").strip()
+
+        # [NEW] 尝试通过 extend_post_id 或者 parent_post_id 获取强绑定的 token
+        from app.services.grok.utils.asset_token_map import AssetTokenMap
+        token_map = await AssetTokenMap.get_instance()
+        bound_token = None
+        if extend_post_id:
+            bound_token = await token_map.get_token(extend_post_id)
+        elif parent_post_id:
+            bound_token = await token_map.get_token(parent_post_id)
+            
+        if bound_token:
+            preferred_token = bound_token
+
         if preferred_token.startswith("sso="):
             preferred_token = preferred_token[4:]
         used_tokens: set[str] = set()
@@ -619,7 +791,7 @@ class VideoService:
                 else:
                     used_tokens.add(preferred_token)
                     logger.warning(
-                        f"Video token routing: preferred token not in pool, fallback to normal routing "
+                        f"Video token routing: preferred bound token not in pool, fallback to normal routing "
                         f"(token={_token_tag(preferred_token)})"
                     )
 
@@ -668,7 +840,22 @@ class VideoService:
 
                 # Generate video.
                 service = VideoService()
-                if parent_post_id:
+                if extend_post_id and video_extension_start_time is not None:
+                    # 视频延长路径
+                    response = await service.generate_extend_video(
+                        token=token,
+                        prompt=prompt,
+                        extend_post_id=extend_post_id,
+                        video_extension_start_time=video_extension_start_time,
+                        original_post_id=original_post_id or "",
+                        file_attachment_id=file_attachment_id or "",
+                        aspect_ratio=aspect_ratio,
+                        video_length=video_length,
+                        resolution=resolution,
+                        preset=preset,
+                        stitch_with_extend=stitch_with_extend,
+                    )
+                elif parent_post_id:
                     response = await service.generate_from_parent_post(
                         token=token,
                         prompt=prompt,
@@ -886,6 +1073,13 @@ class VideoStreamProcessor(BaseProcessor):
                     if progress == 100:
                         video_url = video_resp.get("videoUrl", "")
                         thumbnail_url = video_resp.get("thumbnailImageUrl", "")
+                        
+                        # [NEW] 记录生成的视频对应的 postId 与 token 以备延长
+                        video_post_id = video_resp.get("videoPostId") or self._extract_video_id(video_url)
+                        if video_post_id and self.token:
+                            from app.services.grok.utils.asset_token_map import AssetTokenMap
+                            token_map = await AssetTokenMap.get_instance()
+                            await token_map.save_mapping(video_post_id, self.token)
 
                         if self.think_opened:
                             yield self._sse("\n</think>\n")
@@ -1103,6 +1297,12 @@ class VideoCollectProcessor(BaseProcessor):
                         response_id = resp.get("responseId", "")
                         video_url = video_resp.get("videoUrl", "")
                         thumbnail_url = video_resp.get("thumbnailImageUrl", "")
+                        
+                        # [NEW] 记录生成的视频对应的 postId 与 token 以备延长
+                        if fallback_video_id and self.token:
+                            from app.services.grok.utils.asset_token_map import AssetTokenMap
+                            token_map = await AssetTokenMap.get_instance()
+                            await token_map.save_mapping(fallback_video_id, self.token)
 
                         if video_url:
                             if self.upscale_on_finish:

app/services/grok/utils/asset_token_map.py

@@ -0,0 +1,79 @@
+import json
+import asyncio
+from pathlib import Path
+from typing import Optional, Dict
+
+from app.core.logger import logger
+from app.core.storage import DATA_DIR
+from app.services.grok.utils.locks import _file_lock
+
+
+class AssetTokenMap:
+    """管理 postId 与生成时使用 token 的映射关系"""
+    
+    _instance: Optional["AssetTokenMap"] = None
+    _lock = asyncio.Lock()
+    _file_path = DATA_DIR / "asset_tokens.json"
+
+    def __init__(self):
+        self._cache: Dict[str, str] = {}
+        self._initialized = False
+
+    @classmethod
+    async def get_instance(cls) -> "AssetTokenMap":
+        if cls._instance is None:
+            async with cls._lock:
+                if cls._instance is None:
+                    cls._instance = cls()
+                    await cls._instance._load()
+        return cls._instance
+
+    async def _load(self):
+        """从本地加载映射关系"""
+        if self._initialized:
+            return
+            
+        try:
+            if self._file_path.exists():
+                async with _file_lock("asset_token_map", timeout=5):
+                    content = self._file_path.read_text(encoding="utf-8").strip()
+                    if content:
+                        self._cache = json.loads(content)
+            self._initialized = True
+            logger.info(f"Loaded {len(self._cache)} asset-token mappings from file.")
+        except Exception as e:
+            logger.error(f"Failed to load asset tokens: {e}")
+            self._cache = {}
+
+    async def _save(self):
+        """持久化映射关系"""
+        try:
+            async with _file_lock("asset_token_map", timeout=5):
+                self._file_path.write_text(json.dumps(self._cache, ensure_ascii=False), encoding="utf-8")
+        except Exception as e:
+            logger.error(f"Failed to save asset tokens: {e}")
+
+    async def save_mapping(self, post_id: str, token: str) -> None:
+        """保存 postId -> token 的映射"""
+        if not post_id or not token:
+            return
+            
+        # 标准化 token 格式
+        raw_token = token[4:] if token.startswith("sso=") else token
+        
+        # 只保存需要的键值，防止频繁修改
+        if self._cache.get(post_id) != raw_token:
+            self._cache[post_id] = raw_token
+            asyncio.create_task(self._save())
+            logger.debug(f"Saved token mapping for post_id: {post_id}")
+
+    async def get_token(self, post_id: str) -> Optional[str]:
+        """找回当时生成该 post_id 的 token"""
+        if not post_id:
+            return None
+        
+        token = self._cache.get(post_id)
+        if token:
+            logger.info(f"Retrieved mapped token for extension post_id: {post_id}")
+            return token
+        return None

app/services/reverse/video_upscale.py

@@ -38,7 +38,7 @@ class VideoUpscaleReverse:
                 cookie_token=token,
                 content_type="application/json",
                 origin="https://grok.com",
-                referer="https://grok.com",
+                referer=f"https://grok.com/imagine/post/{video_id}"
             )
 
             # Build payload

app/static/public/js/imagine.js

@@ -998,7 +998,10 @@
       const timestamp = Date.now();
       const seq = meta && meta.sequence ? meta.sequence : 'unknown';
       const ext = mime === 'image/png' ? 'png' : 'jpg';
-      const filename = `imagine_${timestamp}_${seq}.${ext}`;
+      const autoNameParts = ['imagine', String(timestamp)];
+      if (parentPostId) autoNameParts.push(parentPostId.slice(0, 16));
+      autoNameParts.push(String(seq));
+      const filename = `${autoNameParts.join('_')}.${ext}`;
 
       if (useFileSystemAPI && directoryHandle) {
         saveToFileSystem(base64, filename).catch(() => {
@@ -1752,7 +1755,11 @@
           if (!blob) {
             throw new Error('empty blob');
           }
-          const filename = `${prompt.substring(0, 30).replace(/[^a-zA-Z0-9\u4e00-\u9fa5]/g, '_')}_${processed + 1}.png`;
+          const pId = String(item.dataset.parentPostId || '').trim();
+          const nameParts = [prompt.substring(0, 20).replace(/[^a-zA-Z0-9\u4e00-\u9fa5]/g, '_')];
+          if (pId) nameParts.push(pId.slice(0, 16));
+          nameParts.push(String(processed + 1));
+          const filename = `${nameParts.join('_')}.png`;
           imgFolder.file(filename, blob);
           processed++;
 

app/static/public/pages/video.html

@@ -3,7 +3,8 @@
 
 <head>
   <meta charset="UTF-8">
-  <meta name="viewport" content="width=device-width, initial-scale=1, maximum-scale=1, user-scalable=no, viewport-fit=cover, interactive-widget=overlays-content">
+  <meta name="viewport"
+    content="width=device-width, initial-scale=1, maximum-scale=1, user-scalable=no, viewport-fit=cover, interactive-widget=overlays-content">
   <title>Grok2API - Video 视频工作台</title>
   <link rel="icon" href="/static/common/img/favicon/favicon.ico">
   <link rel="manifest" href="/manifest.webmanifest">
@@ -42,80 +43,81 @@
           <div class="video-card">
             <div class="card-title">生成设置</div>
             <div class="settings-grid">
-            <div class="settings-block prompt-block">
-              <label class="field-label" for="promptInput">提示词</label>
-              <textarea id="promptInput" class="geist-input video-textarea" placeholder="例如：街头霓虹雨夜，慢镜头，胶片质感"></textarea>
-            </div>
-            <div id="refDropZone" class="settings-block ref-block">
-              <label class="field-label" for="imageUrlInput">参考图</label>
-              <div class="ref-controls">
-                <input id="imageUrlInput" class="geist-input" placeholder="https://... 或 data:image/...">
+              <div class="settings-block prompt-block">
+                <label class="field-label" for="promptInput">提示词</label>
+                <textarea id="promptInput" class="geist-input video-textarea"
+                  placeholder="例如：街头霓虹雨夜，慢镜头，胶片质感"></textarea>
               </div>
-              <div class="ref-meta">
-                <span id="imageFileName" class="ref-name">未选择文件</span>
-                <span class="ref-drop-hint">支持拖拽/粘贴图片</span>
+              <div id="refDropZone" class="settings-block ref-block">
+                <label class="field-label" for="imageUrlInput">参考图</label>
+                <div class="ref-controls">
+                  <input id="imageUrlInput" class="geist-input" placeholder="https://... 或 data:image/...">
+                </div>
+                <div class="ref-meta">
+                  <span id="imageFileName" class="ref-name">未选择文件</span>
+                  <span class="ref-drop-hint">支持拖拽/粘贴图片</span>
+                </div>
+                <div class="ref-parent-row">
+                  <input id="parentPostInput" class="geist-input" placeholder="粘贴 parentPostId（全局记忆）">
+                  <button id="applyParentBtn" class="geist-button-outline text-xs px-3" type="button">使用ID</button>
+                </div>
+                <div id="referencePreview" class="ref-preview hidden">
+                  <img id="referencePreviewImg" class="ref-preview-image" alt="参考图预览">
+                  <div id="referencePreviewMeta" class="ref-preview-meta"></div>
+                </div>
               </div>
-              <div class="ref-parent-row">
-                <input id="parentPostInput" class="geist-input" placeholder="粘贴 parentPostId（全局记忆）">
-                <button id="applyParentBtn" class="geist-button-outline text-xs px-3" type="button">使用ID</button>
+              <div class="settings-block ratio-block">
+                <label class="field-label" for="ratioSelect">画面比例</label>
+                <select id="ratioSelect" class="geist-input">
+                  <option value="3:2" selected>3:2 横构图</option>
+                  <option value="2:3">2:3 竖构图</option>
+                  <option value="16:9">16:9 宽屏</option>
+                  <option value="9:16">9:16 竖屏</option>
+                  <option value="1:1">1:1 方形</option>
+                </select>
               </div>
-              <div id="referencePreview" class="ref-preview hidden">
-                <img id="referencePreviewImg" class="ref-preview-image" alt="参考图预览">
-                <div id="referencePreviewMeta" class="ref-preview-meta"></div>
+              <div class="settings-block length-block">
+                <label class="field-label" for="lengthSelect">视频时长</label>
+                <select id="lengthSelect" class="geist-input">
+                  <option value="6" selected>6 秒</option>
+                  <option value="10">10 秒</option>
+                  <option value="15">15 秒</option>
+                </select>
+              </div>
+              <div class="settings-block resolution-block">
+                <label class="field-label" for="resolutionSelect">分辨率</label>
+                <select id="resolutionSelect" class="geist-input">
+                  <option value="480p" selected>480p</option>
+                  <option value="720p">720p</option>
+                </select>
+              </div>
+              <div class="settings-block preset-block">
+                <label class="field-label" for="presetSelect">风格预设</label>
+                <select id="presetSelect" class="geist-input">
+                  <option value="normal" selected>Normal</option>
+                  <option value="fun">Fun</option>
+                  <option value="spicy">Spicy</option>
+                  <option value="custom">Custom</option>
+                </select>
+              </div>
+              <div class="settings-block concurrent-block">
+                <label class="field-label" for="concurrentSelect">并发视频数</label>
+                <select id="concurrentSelect" class="geist-input">
+                  <option value="1" selected>1 路</option>
+                  <option value="2">2 路</option>
+                  <option value="3">3 路</option>
+                  <option value="4">4 路</option>
+                </select>
+              </div>
+              <div class="settings-block upload-block">
+                <label class="field-label">&nbsp;</label>
+                <button id="selectImageFileBtn" class="geist-button-outline text-xs px-3" type="button">上传</button>
+                <input id="imageFileInput" class="ref-file-input" type="file" accept="image/*">
+              </div>
+              <div class="settings-block clear-block">
+                <label class="field-label">&nbsp;</label>
+                <button id="clearImageFileBtn" class="geist-button-outline text-xs px-3" type="button">清除</button>
               </div>
-            </div>
-            <div class="settings-block ratio-block">
-              <label class="field-label" for="ratioSelect">画面比例</label>
-              <select id="ratioSelect" class="geist-input">
-                <option value="3:2" selected>3:2 横构图</option>
-                <option value="2:3">2:3 竖构图</option>
-                <option value="16:9">16:9 宽屏</option>
-                <option value="9:16">9:16 竖屏</option>
-                <option value="1:1">1:1 方形</option>
-              </select>
-            </div>
-            <div class="settings-block length-block">
-              <label class="field-label" for="lengthSelect">视频时长</label>
-              <select id="lengthSelect" class="geist-input">
-                <option value="6" selected>6 秒</option>
-                <option value="10">10 秒</option>
-                <option value="15">15 秒</option>
-              </select>
-            </div>
-            <div class="settings-block resolution-block">
-              <label class="field-label" for="resolutionSelect">分辨率</label>
-              <select id="resolutionSelect" class="geist-input">
-                <option value="480p" selected>480p</option>
-                <option value="720p">720p</option>
-              </select>
-            </div>
-            <div class="settings-block preset-block">
-              <label class="field-label" for="presetSelect">风格预设</label>
-              <select id="presetSelect" class="geist-input">
-                <option value="normal" selected>Normal</option>
-                <option value="fun">Fun</option>
-                <option value="spicy">Spicy</option>
-                <option value="custom">Custom</option>
-              </select>
-            </div>
-            <div class="settings-block concurrent-block">
-              <label class="field-label" for="concurrentSelect">并发视频数</label>
-              <select id="concurrentSelect" class="geist-input">
-                <option value="1" selected>1 路</option>
-                <option value="2">2 路</option>
-                <option value="3">3 路</option>
-                <option value="4">4 路</option>
-              </select>
-            </div>
-            <div class="settings-block upload-block">
-              <label class="field-label">&nbsp;</label>
-              <button id="selectImageFileBtn" class="geist-button-outline text-xs px-3" type="button">上传</button>
-              <input id="imageFileInput" class="ref-file-input" type="file" accept="image/*">
-            </div>
-            <div class="settings-block clear-block">
-              <label class="field-label">&nbsp;</label>
-              <button id="clearImageFileBtn" class="geist-button-outline text-xs px-3" type="button">清除</button>
-            </div>
             </div>
             <div class="settings-actions">
               <button id="startBtn" class="geist-button gap-2">
@@ -136,7 +138,7 @@
           </div>
           <div id="editPanel" class="video-card edit-panel">
             <div class="video-preview-header">
-              <div class="card-title">视频工作区</div>
+              <div class="card-title">视频延长工作区</div>
               <div class="preview-actions">
                 <button id="pickCachedVideoBtn" class="geist-button-outline text-xs px-3">选择视频</button>
                 <button id="uploadWorkVideoBtn" class="geist-button-outline text-xs px-3" type="button">上传视频</button>
@@ -147,78 +149,55 @@
             <div id="editHint" class="video-empty">请先选择或生成一个视频，工作区会始终展示当前视频。</div>
             <div id="editBody">
               <div class="meta-grid edit-meta-grid">
-                <div class="meta-item"><div class="meta-label">当前工作视频</div><div id="mergeVideoA" class="meta-value">-</div></div>
-                <div class="meta-item"><div class="meta-label">历史数量</div><div id="historyCount" class="meta-value">0</div></div>
+                <div class="meta-item">
+                  <div class="meta-label">当前工作视频</div>
+                  <div id="mergeVideoA" class="meta-value">-</div>
+                </div>
+                <div class="meta-item">
+                  <div class="meta-label">历史数量</div>
+                  <div id="historyCount" class="meta-value">0</div>
+                </div>
               </div>
               <div class="edit-preview-wrap">
                 <video id="editVideo" controls preload="metadata" playsinline webkit-playsinline></video>
               </div>
               <div class="edit-slider-wrap">
-                <input id="editTimeline" class="timeline-delete-zone" data-delete-side="right" type="range" min="0" max="100000" step="1" value="0">
+                <input id="editTimeline" class="timeline-delete-zone" data-delete-side="right" type="range" min="0"
+                  max="100000" step="1" value="0">
                 <div class="progress-meta">
                   <span id="editTimeText">00:00.000</span>
                   <span id="editDurationText">总时长 -</span>
                 </div>
               </div>
               <div class="meta-grid edit-meta-grid">
-                <div class="meta-item"><div class="meta-label">frame_index</div><div id="editFrameIndex" class="meta-value">-</div></div>
-                <div class="meta-item"><div class="meta-label">timestamp_ms</div><div id="editTimestampMs" class="meta-value">-</div></div>
-                <div class="meta-item"><div class="meta-label">frame_hash</div><div id="editFrameHash" class="meta-value">-</div></div>
-                <div class="meta-item"><div class="meta-label">视频2(B)</div><div id="mergeVideoB" class="meta-value">-</div></div>
+                <div class="meta-item">
+                  <div class="meta-label">frame_index</div>
+                  <div id="editFrameIndex" class="meta-value">-</div>
+                </div>
+                <div class="meta-item">
+                  <div class="meta-label">timestamp_ms</div>
+                  <div id="editTimestampMs" class="meta-value">-</div>
+                </div>
+                <div class="meta-item">
+                  <div class="meta-label">extend_post_id</div>
+                  <div id="editExtendPostId" class="meta-value">-</div>
+                </div>
               </div>
               <div class="settings-block prompt-block edit-prompt-block">
-                <label class="field-label" for="editPromptInput">拼接提示词</label>
-                <textarea id="editPromptInput" class="geist-input video-textarea" placeholder="描述后续要拼接的画面内容"></textarea>
+                <label class="field-label" for="editPromptInput">延长提示词</label>
+                <textarea id="editPromptInput" class="geist-input video-textarea"
+                  placeholder="描述后续要延长的画面内容（留空则自然延续）"></textarea>
               </div>
               <div class="preview-actions splice-action-row">
                 <button id="spliceBtn" class="geist-button gap-2">
                   <svg width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"
                     stroke-linecap="round" stroke-linejoin="round">
-                    <path d="M11 4H6a2 2 0 0 0-2 2v5" />
-                    <path d="M13 20h5a2 2 0 0 0 2-2v-5" />
-                    <path d="M20 6l-8 8" />
-                    <path d="M4 18l8-8" />
+                    <path d="M5 12h14" />
+                    <path d="M12 5v14" />
                   </svg>
-                  <span>拼接视频</span>
+                  <span>延长视频</span>
                 </button>
               </div>
-              <div class="video-card manual-merge-card">
-                <div class="video-preview-header">
-                  <div class="card-title">手动两段拼接</div>
-                  <div class="preview-actions">
-                    <button id="pickMergeVideoBtn" class="geist-button-outline text-xs px-3" type="button">选择视频2</button>
-                    <button id="directMergeBtn" class="geist-button-outline text-xs px-3" type="button">手动拼接</button>
-                  </div>
-                </div>
-                <div class="manual-merge-grid">
-                  <div class="manual-merge-col">
-                    <div class="meta-label">视频1拼接帧（向左保留）</div>
-                    <div class="edit-preview-wrap">
-                      <video id="mergeVideoPreviewA" controls preload="metadata" playsinline webkit-playsinline></video>
-                    </div>
-                    <div class="edit-slider-wrap">
-                      <input id="mergeTimelineA" class="timeline-delete-zone" data-delete-side="right" type="range" min="0" max="100000" step="1" value="0">
-                      <div class="progress-meta">
-                        <span id="mergeTimeTextA">00:00.000</span>
-                        <span id="mergeDurationA">总时长 -</span>
-                      </div>
-                    </div>
-                  </div>
-                  <div class="manual-merge-col">
-                    <div class="meta-label">视频2拼接帧（向右保留）</div>
-                    <div class="edit-preview-wrap">
-                      <video id="mergeVideoPreviewB" controls preload="metadata" playsinline webkit-playsinline></video>
-                    </div>
-                    <div class="edit-slider-wrap">
-                      <input id="mergeTimelineB" class="timeline-delete-zone" data-delete-side="left" type="range" min="0" max="100000" step="1" value="0">
-                      <div class="progress-meta">
-                        <span id="mergeTimeTextB">00:00.000</span>
-                        <span id="mergeDurationB">总时长 -</span>
-                      </div>
-                    </div>
-                  </div>
-                </div>
-              </div>
             </div>
           </div>
         </div>
@@ -266,7 +245,7 @@
             <div class="panel-head">
               <div class="card-title">历史视频</div>
             </div>
-            <div id="videoEmpty" class="video-empty">生成视频与拼接视频都会在这里保存，点击可替换到工作区。</div>
+            <div id="videoEmpty" class="video-empty">生成视频与延长视频都会在这里保存，点击可替换到工作区。</div>
             <div id="videoStage" class="video-stage hidden"></div>
           </aside>
         </aside>
@@ -298,9 +277,7 @@
   <script src="/static/common/js/toast.js"></script>
   <script src="/static/common/js/prompt-enhancer.js?v=0.1.15"></script>
   <script src="/static/common/js/pwa.js?v=0.1.1"></script>
-  <script src="/static/vendor/ffmpeg/ffmpeg.js?v=0.1.0"></script>
-  <script src="/static/vendor/ffmpeg/ffmpeg-util.js?v=0.1.0"></script>
-  <script src="/static/public/js/video.js?v=0.5.42"></script>
+  <script src="/static/public/js/video.js?v=0.6.0"></script>
 </body>
 
-</html>
+</html>