The generated video lacks sufficient clarity

Even when using high-quality input images to generate the video, the output lacks sufficient clarity.

After reviewing the underlying bucket_tools.py code, I noticed that the input images are resized, and the available width/height options are quite limited. This results in processed images losing clarity, which in turn makes the generated video less sharp.

I'd like to ask: Are we restricted to only these preset options? Is it possible to adjust the parameters to allow higher-resolution video generation?

bucket_options = {
    640: [
        (416, 960),
        (448, 864),
        (480, 832),
        (512, 768),
        (544, 704),
        (576, 672),
        (608, 640),
        (640, 608),
        (672, 576),
        (704, 544),
        (768, 512),
        (832, 480),
        (864, 448),
        (960, 416),
    ],
}


def find_nearest_bucket(h, w, resolution=640):
    min_metric = float('inf')
    best_bucket = None
    for (bucket_h, bucket_w) in bucket_options[resolution]:
        metric = abs(h * bucket_w - w * bucket_h)
        if metric <= min_metric:
            min_metric = metric
            best_bucket = (bucket_h, bucket_w)
    return best_bucket

Start by reading bucket_tools.py, focusing on bucket_options and find_nearest_bucket. Trace where the selected resolution and buckets are used in video generation, then determine whether higher-resolution options are supported by the surrounding pipeline. Done means the supported resolution limits and any required changes are clearly established and covered by relevant validation.