Update handler.py
Browse files- handler.py +70 -6
handler.py
CHANGED
|
@@ -69,6 +69,70 @@ apply_dirty_hack_to_patch_file_extensions_and_bypass_filter("/repository")
|
|
| 69 |
#logger.info("💡 Printing directory structure of ""/repository"":")
|
| 70 |
#print_directory_structure("/repository")
|
| 71 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 72 |
@dataclass
|
| 73 |
class GenerationConfig:
|
| 74 |
"""Configuration for video generation"""
|
|
@@ -339,12 +403,12 @@ class EndpointHandler:
|
|
| 339 |
|
| 340 |
# Check if image-to-video generation is requested
|
| 341 |
if input_image:
|
| 342 |
-
|
| 343 |
-
|
| 344 |
-
|
| 345 |
-
|
| 346 |
-
|
| 347 |
-
generation_kwargs["image"] =
|
| 348 |
frames = self.image_to_video(**generation_kwargs).frames
|
| 349 |
else:
|
| 350 |
frames = self.text_to_video(**generation_kwargs).frames
|
|
|
|
| 69 |
#logger.info("💡 Printing directory structure of ""/repository"":")
|
| 70 |
#print_directory_structure("/repository")
|
| 71 |
|
| 72 |
+
|
| 73 |
+
def process_input_image(image_data: str, target_width: int, target_height: int) -> Image.Image:
|
| 74 |
+
"""
|
| 75 |
+
Process input image from base64, resize and crop to target dimensions
|
| 76 |
+
|
| 77 |
+
Args:
|
| 78 |
+
image_data: Base64 encoded image data
|
| 79 |
+
target_width: Desired width
|
| 80 |
+
target_height: Desired height
|
| 81 |
+
|
| 82 |
+
Returns:
|
| 83 |
+
Processed PIL Image
|
| 84 |
+
"""
|
| 85 |
+
try:
|
| 86 |
+
# Handle data URI format
|
| 87 |
+
if image_data.startswith('data:'):
|
| 88 |
+
image_data = image_data.split(',', 1)[1]
|
| 89 |
+
|
| 90 |
+
# Decode base64
|
| 91 |
+
image_bytes = base64.b64decode(image_data)
|
| 92 |
+
image = Image.open(io.BytesIO(image_bytes))
|
| 93 |
+
|
| 94 |
+
# Convert to RGB if necessary
|
| 95 |
+
if image.mode not in ('RGB', 'RGBA'):
|
| 96 |
+
image = image.convert('RGB')
|
| 97 |
+
elif image.mode == 'RGBA':
|
| 98 |
+
# Handle transparency by compositing on white background
|
| 99 |
+
background = Image.new('RGB', image.size, (255, 255, 255))
|
| 100 |
+
background.paste(image, mask=image.split()[3])
|
| 101 |
+
image = background
|
| 102 |
+
|
| 103 |
+
# Calculate target aspect ratio
|
| 104 |
+
target_aspect = target_width / target_height
|
| 105 |
+
|
| 106 |
+
# Get current dimensions
|
| 107 |
+
orig_width, orig_height = image.size
|
| 108 |
+
orig_aspect = orig_width / orig_height
|
| 109 |
+
|
| 110 |
+
# Calculate dimensions for resizing
|
| 111 |
+
if orig_aspect > target_aspect:
|
| 112 |
+
# Image is wider than target
|
| 113 |
+
new_height = target_height
|
| 114 |
+
new_width = int(target_height * orig_aspect)
|
| 115 |
+
else:
|
| 116 |
+
# Image is taller than target
|
| 117 |
+
new_width = target_width
|
| 118 |
+
new_height = int(target_width / orig_aspect)
|
| 119 |
+
|
| 120 |
+
# Resize image
|
| 121 |
+
image = image.resize((new_width, new_height), Image.Resampling.LANCZOS)
|
| 122 |
+
|
| 123 |
+
# Center crop to target dimensions
|
| 124 |
+
left = (new_width - target_width) // 2
|
| 125 |
+
top = (new_height - target_height) // 2
|
| 126 |
+
right = left + target_width
|
| 127 |
+
bottom = top + target_height
|
| 128 |
+
|
| 129 |
+
image = image.crop((left, top, right, bottom))
|
| 130 |
+
|
| 131 |
+
return image
|
| 132 |
+
|
| 133 |
+
except Exception as e:
|
| 134 |
+
raise ValueError(f"Failed to process input image: {str(e)}")
|
| 135 |
+
|
| 136 |
@dataclass
|
| 137 |
class GenerationConfig:
|
| 138 |
"""Configuration for video generation"""
|
|
|
|
| 403 |
|
| 404 |
# Check if image-to-video generation is requested
|
| 405 |
if input_image:
|
| 406 |
+
processed_image = process_input_image(
|
| 407 |
+
input_image,
|
| 408 |
+
config.width,
|
| 409 |
+
config.height
|
| 410 |
+
)
|
| 411 |
+
generation_kwargs["image"] = processed_image
|
| 412 |
frames = self.image_to_video(**generation_kwargs).frames
|
| 413 |
else:
|
| 414 |
frames = self.text_to_video(**generation_kwargs).frames
|