diff --git a/entity/ffmpeg.py b/entity/ffmpeg.py
index 51a9c53..3e864a0 100644
--- a/entity/ffmpeg.py
+++ b/entity/ffmpeg.py
@@ -299,11 +299,42 @@ class FfmpegTask(object):
                     if zoom_factor == 1:
                         continue
                     effect_index += 1
+                    
+                    # 获取缩放中心点（从pos_json或使用默认中心）
+                    center_x = "iw/2"
+                    center_y = "ih/2"
+                    pos_json_str = self.ext_data.get('posJson', '{}')
+                    try:
+                        pos_json = json.loads(pos_json_str) if pos_json_str != '{}' else {}
+                        if pos_json:
+                            _f_x = pos_json.get('ltX', 0)
+                            _f_x2 = pos_json.get('rbX', 0)
+                            _f_y = pos_json.get('ltY', 0)
+                            _f_y2 = pos_json.get('rbY', 0)
+                            _v_w = pos_json.get('imgWidth', 1)
+                            _v_h = pos_json.get('imgHeight', 1)
+                            if _v_w > 0 and _v_h > 0:
+                                # 计算坐标系统中的中心点
+                                center_x_ratio = (_f_x + _f_x2) / (2 * _v_w)
+                                center_y_ratio = (_f_y + _f_y2) / (2 * _v_h)
+                                # 转换为视频坐标系统
+                                center_x = f"iw*{center_x_ratio:.6f}"
+                                center_y = f"ih*{center_y_ratio:.6f}"
+                    except Exception as e:
+                        # 解析失败使用默认中心
+                        pass
+                    
                     if duration == 0:
-                        filter_args.append(f"{video_output_str}trim=start={start_time},scale=iw*{zoom_factor}:ih*{zoom_factor},crop=iw/{zoom_factor}:ih/{zoom_factor}:(iw-iw/{zoom_factor})/2:(ih-ih/{zoom_factor})/2,setpts=PTS-STARTPTS[v_eff{effect_index}]")
+                        # 静态缩放（整个视频时长）
+                        x_expr = f"({center_x})-(ow*zoom)/2"
+                        y_expr = f"({center_y})-(oh*zoom)/2"
+                        filter_args.append(f"{video_output_str}trim=start={start_time},zoompan=z={zoom_factor}:x={x_expr}:y={y_expr}:d=1[v_eff{effect_index}]")
                     else:
-                        zoom_expr = f"if(between(t,{start_time},{start_time + duration}),{zoom_factor},1)"
-                        filter_args.append(f"{video_output_str}scale=iw*({zoom_expr}):ih*({zoom_expr}),crop=iw:ih:(iw-iw)/2:(ih-ih)/2[v_eff{effect_index}]")
+                        # 动态缩放（指定时间段内）
+                        zoom_expr = f"if(between(t\\,{start_time}\\,{start_time + duration})\\,{zoom_factor}\\,1)"
+                        x_expr = f"({center_x})-(ow*zoom)/2"
+                        y_expr = f"({center_y})-(oh*zoom)/2"
+                        filter_args.append(f"{video_output_str}zoompan=z={zoom_expr}:x={x_expr}:y={y_expr}:d=1[v_eff{effect_index}]")
                     video_output_str = f"[v_eff{effect_index}]"
                 elif effect.startswith("skip:"):
                     param = effect.split(":", 2)[1]
@@ -312,7 +343,7 @@ class FfmpegTask(object):
                     skip_seconds = float(param)
                     if skip_seconds > 0:
                         effect_index += 1
-                        filter_args.append(f"{video_output_str}trim=start={skip_seconds},setpts=PTS-STARTPTS[v_eff{effect_index}]")
+                        filter_args.append(f"{video_output_str}trim=start={skip_seconds}[v_eff{effect_index}]")
                         video_output_str = f"[v_eff{effect_index}]"
                 elif effect.startswith("tail:"):
                     param = effect.split(":", 2)[1]
@@ -324,7 +355,7 @@ class FfmpegTask(object):
                         # 首先获取视频总时长，然后计算开始时间
                         # 使用reverse+trim+reverse的方法来精确获取最后N秒
                         filter_args.append(f"{video_output_str}reverse[v_rev{effect_index}]")
-                        filter_args.append(f"[v_rev{effect_index}]trim=duration={tail_seconds},setpts=PTS-STARTPTS[v_trim{effect_index}]")
+                        filter_args.append(f"[v_rev{effect_index}]trim=duration={tail_seconds}[v_trim{effect_index}]")
                         filter_args.append(f"[v_trim{effect_index}]reverse[v_eff{effect_index}]")
                         video_output_str = f"[v_eff{effect_index}]"
                 ...