app.py

import gradio as gr
import spaces
import torch
import os
from glob import glob
from diffusers import StableVideoDiffusionPipeline
from diffusers.utils import export_to_video
from PIL import Image

output_folder = "outputs"

pipe = StableVideoDiffusionPipeline.from_pretrained(
    "stabilityai/stable-video-diffusion-img2vid-xt", variant="fp16"
).to("cuda")

@spaces.GPU(duration=480)
def sample(
    image: Image,
    width: int = 1024,
    height: int = 576,
    motion_bucket_id: int = 127,
    fps_id: int = 30,
):
    width = int(width)
    height = int(height)
    img = image.resize((width, height))

    os.makedirs(output_folder, exist_ok=True)
    base_count = len(glob(os.path.join(output_folder, "*.mp4")))
    video_path = os.path.join(output_folder, f"{base_count:06d}.mp4")

    frames = pipe(
        img,
        decode_chunk_size=3,
        generator=None,
        motion_bucket_id=motion_bucket_id,
        noise_aug_strength=0.1,
        num_frames=25,
    ).frames[0]

    export_to_video(frames, video_path, fps=fps_id)
    return video_path

with gr.Blocks() as demo:
    with gr.Row():
        image = gr.Image(label="Upload your image", type="pil")
        video = gr.Video()

    with gr.Column():
        generate_btn = gr.Button("Generate")

        with gr.Accordion("Advanced options", open=False):
            width = gr.Number(label="Width", value=1024, minimum=1)
            height = gr.Number(label="Height", value=576, minimum=1)
            motion_bucket_id = gr.Slider(
                label="Motion bucket id",
                info="Controls how much motion to add/remove from the image",
                value=60,
                minimum=1,
                maximum=255,
            )
            fps_id = gr.Slider(
                label="Frames per second",
                info="Video length will be 25 frames.",
                value=30,
                minimum=5,
                step=2,
                maximum=60,
            )

    image.upload(fn=lambda img: img, inputs=image, outputs=image, queue=False)
    generate_btn.click(
        fn=sample,
        inputs=[image, width, height, motion_bucket_id, fps_id],
        outputs=[video],
        api_name="video",
    )

if __name__ == "__main__":
    demo.queue(max_size=20, api_open=False)
    demo.launch(show_api=False)