feat: preserve original video fps by reusing the input fps in AniDoc generation and final resized output

gradio_app.py

@@ -188,7 +188,7 @@ def cleanup_runtime_models(models) -> None:
                 pass
 
 # -----------------------------------------------------------------------------
-# Video post-processing
+# Video metadata / post-processing
 # -----------------------------------------------------------------------------
 
 def get_video_dimensions(video_path: str) -> tuple[int, int]:
@@ -206,11 +206,26 @@ def get_video_dimensions(video_path: str) -> tuple[int, int]:
     return width, height
 
 
+def get_video_fps(video_path: str) -> float:
+    cap = cv2.VideoCapture(video_path)
+    if not cap.isOpened():
+        raise ValueError(f"Unable to open input video: {video_path}")
+
+    fps = cap.get(cv2.CAP_PROP_FPS)
+    cap.release()
+
+    if fps <= 0:
+        fps = DEFAULT_FPS
+
+    return round(float(fps), 2)
+
+
 def resize_output_video_to_input_size(
     output_video_path: str,
     input_video_path: str,
 ) -> str:
     target_width, target_height = get_video_dimensions(input_video_path)
+    target_fps = get_video_fps(input_video_path)
     resized_output_path = os.path.splitext(output_video_path)[0] + "_resized.mp4"
 
     cmd = [
@@ -218,6 +233,7 @@ def resize_output_video_to_input_size(
         "-y",
         "-i", output_video_path,
         "-vf", f"scale={target_width}:{target_height}",
+        "-r", str(target_fps),
         "-c:v", "libx264",
         "-pix_fmt", "yuv420p",
         resized_output_path,
@@ -437,6 +453,10 @@ def run_anidoc(
         max_points=max_points,
     )
 
+    input_fps = DEFAULT_FPS
+    if control_path.lower().endswith(".mp4"):
+        input_fps = get_video_fps(control_path)
+
     generator = torch.manual_seed(seed)
 
     with torch.inference_mode():
@@ -448,7 +468,7 @@ def run_anidoc(
             num_frames=num_frames,
             decode_chunk_size=DEFAULT_DECODE_CHUNK_SIZE,
             motion_bucket_id=DEFAULT_MOTION_BUCKET_ID,
-            fps=DEFAULT_FPS,
+            fps=input_fps,
             noise_aug_strength=DEFAULT_NOISE_AUG,
             generator=generator,
         ).frames[0]