Update app.py

app.py

@@ -23,10 +23,11 @@ def process_video_with_mask(video_path, mask_path, text="Joker"):
     width = int(cap_video.get(cv2.CAP_PROP_FRAME_WIDTH))
     height = int(cap_video.get(cv2.CAP_PROP_FRAME_HEIGHT))
     
-    # 第一阶段：寻找第一个接近水平的角度
+    # 初始化固定参数
+    fixed_font_size = None
     fixed_angle = None
-    current_pos = cap_video.get(cv2.CAP_PROP_POS_FRAMES)
     
+    # 第一阶段：寻找第一个接近水平的角度和计算字体大小
     while fixed_angle is None:
         ret_video, frame_video = cap_video.read()
         ret_mask, frame_mask = cap_mask.read()
@@ -50,16 +51,21 @@ def process_video_with_mask(video_path, mask_path, text="Joker"):
             
             # 获取旋转矩形
             rect = cv2.minAreaRect(max_contour)
+            size = rect[1]
             current_angle = rect[2]
             
+            # 计算基于mask大小的字体大小
+            if fixed_font_size is None:
+                fixed_font_size = int(min(size[0] * 0.8 / max(1, len(text)), size[1] * 0.8))
+            
             # 检查是否接近水平
             if is_horizontal_angle(current_angle):
                 fixed_angle = current_angle
     
     # 如果没有找到接近水平的角度，使用第一个找到的角度
     if fixed_angle is None:
-        cap_video.set(cv2.CAP_PROP_POS_FRAMES, current_pos)
-        cap_mask.set(cv2.CAP_PROP_POS_FRAMES, current_pos)
+        cap_video.set(cv2.CAP_PROP_POS_FRAMES, 0)
+        cap_mask.set(cv2.CAP_PROP_POS_FRAMES, 0)
         
         ret_video, frame_video = cap_video.read()
         ret_mask, frame_mask = cap_mask.read()
@@ -88,7 +94,7 @@ def process_video_with_mask(video_path, mask_path, text="Joker"):
     fourcc = cv2.VideoWriter_fourcc(*'mp4v')
     out = cv2.VideoWriter(output_path, fourcc, fps, (width, height))
     
-    # 第二阶段：使用固定角度处理视频
+    # 第二阶段：使用固定角度和固定字体大小处理视频
     while True:
         ret_video, frame_video = cap_video.read()
         ret_mask, frame_mask = cap_mask.read()
@@ -116,10 +122,6 @@ def process_video_with_mask(video_path, mask_path, text="Joker"):
             
             # 获取旋转矩形
             rect = cv2.minAreaRect(max_contour)
-            box = cv2.boxPoints(rect)
-            box = np.int0(box)
-            
-            # 计算中心点和尺寸
             center = rect[0]
             size = rect[1]
             
@@ -128,9 +130,7 @@ def process_video_with_mask(video_path, mask_path, text="Joker"):
             text_draw = ImageDraw.Draw(text_image)
             
             try:
-                # 自动调整字体大小
-                font_size = int(min(size[0] * 0.8 / max(1, len(text)), size[1] * 0.8))
-                font = ImageFont.truetype("华文琥珀.ttf", font_size)
+                font = ImageFont.truetype("华文琥珀.ttf", fixed_font_size)
             except:
                 font = ImageFont.load_default()
             
@@ -185,7 +185,9 @@ with gr.Blocks() as demo:
 
     gr.Examples(
         [
-            ["examples/20250511-013442_3331156457_.mp4", "examples/20250511-013442_3331156457__temp.mp4_alpha_temp.mp4", "今天天气怎么样？"]
+            ["examples/20250511-013442_3331156457_.mp4", "examples/20250511-013442_3331156457__temp.mp4_alpha_temp.mp4", "今天天气怎么样？"],
+            ["examples/20250510-123229_332745243_.mp4", "examples/20250510-123229_332745243__temp.mp4_alpha_temp.mp4", "今天天气怎么样？"],
+            ["examples/20250510-121949_3109246688_.mp4", "examples/20250510-121949_3109246688__temp.mp4_alpha_temp.mp4", "今天天气怎么样？"],
         ],
         [video_input, mask_input, text_input]
     )