custom-diffusion 0.1.9

Custom Diffusion: Creating Video from Frame Using Diffusion

Installation

pip install custom_diffusion

Usage

# Importing the required libraries

from custom_diffusion.utils.data_utils import load_images_from_folder
from custom_diffusion import StableDiffusionControlNetGenerator
from custom_diffusion.utils.video_utils import convert_images_to_video
from custom_diffusion.demo import video_pipeline

# Creating a video from a video file
frames_path = video_pipeline(
    video_path="test.mp4",
    output_path="output.mp4",
    start_time=0,
    end_time=5,
    frame_rate=1,
)

# Creating a video from a folder of images
images_list = load_images_from_folder(frames_path)

prompt = "a anime boy"
negative_prompt = "bad"

list_prompt = [prompt] * len(images_list)
list_negative_prompt = [negative_prompt] * len(images_list)

# Generating images from a list of images
generator = StableDiffusionControlNetGenerator()

generated_image_list = generator.generate_image(
    stable_model_path="andite/anything-v4.0",
    controlnet_model_path="lllyasviel/control_v11p_sd15_canny",
    scheduler_name="EulerAncestralDiscrete",
    images_list=images_list,
    prompt=list_prompt,
    negative_prompt=list_negative_prompt,
    height=512,
    width=512,
    guess_mode=False,
    num_images_per_prompt=1,
    num_inference_steps=30,
    guidance_scale=7.0,
    controlnet_conditioning_scale=1.0,
    generator_seed=0,
    preprocess_type="Canny",
    resize_type="center_crop_and_resize",
    crop_size=512,
)

# Converting the generated images to a video
frame2video = convert_images_to_video(
    image_list=generated_image_list,
    output_file="output.mp4",
    frame_rate=5,
)