app.py

import spaces
import gradio as gr
import torch
import numpy as np
from diffusers import WanImageToVideoPipeline
from diffusers.utils import export_to_video

model_id = "Wan-AI/Wan2.2-I2V-A14B-Diffusers"

print(f"Using video Model: {model_id}")

dtype = torch.bfloat16
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Load pipeline
pipe = WanImageToVideoPipeline.from_pretrained(
    model_id,
    torch_dtype=dtype
)

pipe.to(device)
print(f"Model Loaded in {device}")
pipe.vae.enable_tiling()

# ================================
# Image Preparation
# ================================

def prepare_vertical_image(pipe, image, base_width=384, base_height=672):

    mod_value = (
        pipe.vae_scale_factor_spatial *
        pipe.transformer.config.patch_size[1]
    )

    final_width = (base_width // mod_value) * mod_value
    final_height = (base_height // mod_value) * mod_value

    resized_image = image.resize((final_width, final_height))

    return resized_image, final_width, final_height


# ================================
# Video Generation
# ================================

@spaces.GPU(size="xlarge",duration=180)
def generate_video(input_image, prompt, negative_prompt, progress=gr.Progress(track_tqdm=True)):

    if input_image is None:
        return None

    image, width, height = prepare_vertical_image(pipe, input_image)

    print(f"Generating vertical video {width}x{height}")

    video_frames = pipe(
        image=image,
        prompt=prompt,
        negative_prompt=negative_prompt,
        height=height,
        width=width,
        num_frames=161,   # FIXED
        guidance_scale=5.0,
        num_inference_steps=15
    ).frames[0]

    output_path = "vertical_output.mp4"

    export_to_video(video_frames, output_path, fps=16)

    return output_path


# ================================
# Gradio UI
# ================================

with gr.Blocks(title="Wan 2.2 Vertical I2V") as demo:

    gr.Markdown("# 🎬 Wan 2.2 Image → Video Generator")
    gr.Markdown("Generate **10-second Vertical (9:16) AI Videos**")

    with gr.Row():

        # LEFT SIDE (INPUTS)
        with gr.Column(scale=1):

            input_image = gr.Image(
                type="pil",
                label="Upload Image"
            )

            prompt = gr.Textbox(
                label="Prompt",
                placeholder="Describe motion, camera movement..."
            )

            negative_prompt = gr.Textbox(
                label="Negative Prompt",
                value="blurry, low quality, distorted, static"
            )

            generate_btn = gr.Button("Generate Video", variant="primary")

        # RIGHT SIDE (OUTPUT)
        with gr.Column(scale=1):

            output_video = gr.Video(
                label="Generated Video"
            )

    generate_btn.click(
        generate_video,
        inputs=[input_image, prompt, negative_prompt],
        outputs=output_video
    )

demo.launch(server_name="0.0.0.0", server_port=7860)