app.py

import gradio as gr
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

# Use CogVideo, a publicly available text-to-video model on Hugging Face.
# (Replace 'THUDM/CogVideo-5M' with the correct ID if needed.)
video_model = pipeline(Tasks.text_to_video_synthesis, model='THUDM/CogVideoX-5b')

def generate_pokemon_anime_video(prompt, style, duration, image):
    """
    Generate a Pokémon anime–themed video.
    The prompt is enriched with style and context so that the model "imagines"
    iconic Pokémon scenes (e.g., Ash, Pikachu, Team Rocket).
    """
    full_prompt = (
        f"{prompt}, in {style} style. "
        "Include iconic Pokémon elements like Ash, Pikachu, and Team Rocket."
    )
    
    # Prepare inputs for the model.
    # (If the model doesn't support image conditioning, the 'image' key may be ignored.)
    inputs = {'text': full_prompt, 'duration': duration}
    if image:
        inputs['image'] = image

    result = video_model(inputs)
    return result["output_video"]

# Build the Gradio UI with dropdown, slider, and file upload.
with gr.Blocks() as iface:
    gr.Markdown("# 🎥 PokeVidGen AI")
    gr.Markdown(
        "Generate Pokémon anime shorts with AI! "
        "Enter a scene prompt, select an animation style, set the video duration, and optionally upload an image."
    )
    with gr.Row():
        prompt = gr.Textbox(
            label="Enter Pokémon Scene",
            placeholder="Ash battles Team Rocket with Pikachu's Thunderbolt"
        )
        style = gr.Dropdown(
            ["Anime Classic", "Modern 3D", "Cartoon"],
            label="Animation Style",
            value="Anime Classic"
        )
    duration = gr.Slider(1, 10, step=1, label="Video Duration (Seconds)", value=5)
    image = gr.Image(label="Upload an Image (Optional)", type="filepath")
    generate_btn = gr.Button("Generate Pokémon Anime Video")
    output_video = gr.Video(label="Generated Video")

    generate_btn.click(generate_pokemon_anime_video,
                       inputs=[prompt, style, duration, image],
                       outputs=output_video)

iface.launch()