Update app.py

app.py

@@ -206,9 +206,20 @@ def interpolate_video_frames(
     output_fps=hardcoded_fps,
     desired_duration=hardcoded_duration_sec,
     original_duration=hardcoded_duration_sec,
-    use_cuda=True, # this requires FFmpeg to have been compiled with CUDA support (to try - I'm not sure the Hugging Face image has that by default)
+    output_width=1024,
+    output_height=576,
+    use_cuda=False, # this requires FFmpeg to have been compiled with CUDA support (to try - I'm not sure the Hugging Face image has that by default)
     verbose=False):
+        
     scale_factor = desired_duration / original_duration
+
+    filters = []
+
+    # Scaling if dimensions are provided
+    if output_width and output_height:
+        filters.append(f'scale={output_width}:{output_height}')
+
+
     # note: from all fact, it looks like using a small macroblock is important for us,
     # since the video resolution is very small (usually 512x288px)
     interpolation_filter = f'minterpolate=mi_mode=mci:mc_mode=obmc:me=hexbs:vsbmc=1:mb_size=4:fps={output_fps}:scd=none,setpts={scale_factor}*PTS'
@@ -221,6 +232,13 @@ def interpolate_video_frames(
     #- `scd=none`: Disables scene change detection entirely. 
     #- `setpts={scale_factor}*PTS`: Adjusts for the stretching of the video duration.
 
+    # Frame interpolation setup
+    filters.append(interpolation_filter)
+
+    # Combine all filters into a single filter complex
+    filter_complex = ','.join(filters)
+
+
     cmd = [
         'ffmpeg',
         '-i', input_file_path,
@@ -230,7 +248,7 @@ def interpolate_video_frames(
         cmd.extend(['-hwaccel', 'cuda', '-hwaccel_output_format', 'cuda'])
                
     cmd.extend([
-        '-filter:v', interpolation_filter,
+        '-filter:v', filter_complex,
         '-r', str(output_fps),
         output_file_path
     ])