Update handler.py

handler.py

@@ -22,9 +22,9 @@ logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
 # Constraints
-MAX_WIDTH = 1280
-MAX_HEIGHT = 720
-MAX_FRAMES = 257
+MAX_LARGE_SIDE = 1280
+MAX_SMALL_SIDE = 720
+MAX_FRAMES = (8 x 21) + 1 # visual glitches appear after about 169 frames, so we cap it
 
 # this is only a temporary solution (famous last words)
 def apply_dirty_hack_to_patch_file_extensions_and_bypass_filter(directory):
@@ -78,12 +78,13 @@ class GenerationConfig:
     negative_prompt: str = "saturated, overlit, worst quality, inconsistent motion, blurry, jittery, distorted, cropped, watermarked, watermark, logo, subtitle, subtitles, lowres"
 
     # video model settings (will be used during generation of the initial raw video clip)
-    width: int = 768
-    height: int = 512
+    width: int = 768 # max is 1280 but we use a lower value
+    height: int = 416 # max is 720 but we use a lower value
 
     # users may tend to always set this to the max, to get as much useable content as possible (which is MAX_FRAMES ie. 257).
     # The value must be a multiple of 8, plus 1 frame.
-    num_frames: int = 129
+    # visual glitches appear after about 169 frames, so we don't need more actually
+    num_frames: int = (8 * 14) + 1
     
     guidance_scale: float = 7.5
     num_inference_steps: int = 50
@@ -92,7 +93,7 @@ class GenerationConfig:
     seed: int = -1  # -1 means random seed
 
     # varnish settings (will be used for post-processing after the raw video clip has been generated
-    fps: int = 24 # FPS of the final video (only applied at the the very end, when converting to mp4)
+    fps: int = 30 # FPS of the final video (only applied at the the very end, when converting to mp4)
     double_num_frames: bool = True # if True, the number of frames will be multiplied by 2 using RIFE
     super_resolution: bool = True # if True, the resolution will be multiplied by 2 using Real_ESRGAN
     
@@ -105,19 +106,31 @@ class GenerationConfig:
 
     def validate_and_adjust(self) -> 'GenerationConfig':
         """Validate and adjust parameters to meet constraints"""
-        # Round dimensions to nearest multiple of 32
-        self.width = max(32, min(MAX_WIDTH, round(self.width / 32) * 32))
-        self.height = max(32, min(MAX_HEIGHT, round(self.height / 32) * 32))
+        # First check if it's one of our explicitly allowed resolutions
+        if not ((self.width == MAX_LARGE_SIDE and self.height == MAX_SMALL_SIDE) or 
+                (self.width == MAX_SMALL_SIDE and self.height == MAX_LARGE_SIDE)):
+            # For other resolutions, ensure total pixels don't exceed max
+            MAX_TOTAL_PIXELS = (MAX_SMALL_SIDE * MAX_LARGE_SIDE- # or 921600 = 1280 * 720
+            
+            # If total pixels exceed maximum, scale down proportionally
+            total_pixels = self.width * self.height
+            if total_pixels > MAX_TOTAL_PIXELS:
+                scale = (MAX_TOTAL_PIXELS / total_pixels) ** 0.5
+                self.width = max(32, min(MAX_LARGE_SIDE, round(self.width * scale / 32) * 32))
+                self.height = max(32, min(MAX_LARGE_SIDE, round(self.height * scale / 32) * 32))
+            else:
+                # Round dimensions to nearest multiple of 32
+                self.width = max(32, min(MAX_LARGE_SIDE, round(self.width / 32) * 32))
+                self.height = max(32, min(MAX_LARGE_SIDE, round(self.height / 32) * 32))
         
         # Adjust number of frames to be in format 8k + 1
         k = (self.num_frames - 1) // 8
-        num_frames = min((k * 8) + 1, MAX_FRAMES)
-
-
+        self.num_frames = min((k * 8) + 1, MAX_FRAMES)
+    
         # Set random seed if not specified
         if self.seed == -1:
             self.seed = random.randint(0, 2**32 - 1)
-
+    
         return self
 
 class EndpointHandler: