import torch
import gradio as gr
from diffusers import AnimateDiffPipeline, MotionAdapter, DPMSolverMultistepScheduler, AutoencoderKL, SparseControlNetModel
from diffusers.utils import export_to_gif, load_image
from transformers import pipeline

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

# 한글-영어 번역 모델 로드
translator = pipeline("translation", model="Helsinki-NLP/opus-mt-ko-en")

def translate_korean_to_english(text):
    if any('\u3131' <= char <= '\u3163' or '\uac00' <= char <= '\ud7a3' for char in text):
        translated = translator(text)[0]['translation_text']
        return translated
    return text

def generate_video(prompt, negative_prompt, num_inference_steps, conditioning_frame_indices, controlnet_conditioning_scale):
    prompt = translate_korean_to_english(prompt)
    negative_prompt = translate_korean_to_english(negative_prompt)

    motion_adapter = MotionAdapter.from_pretrained("guoyww/animatediff-motion-adapter-v1-5-3", torch_dtype=torch.float16).to(device)
    controlnet = SparseControlNetModel.from_pretrained("guoyww/animatediff-sparsectrl-scribble", torch_dtype=torch.float16).to(device)
    vae = AutoencoderKL.from_pretrained("stabilityai/sd-vae-ft-mse", torch_dtype=torch.float16).to(device)
    
    pipe = AnimateDiffPipeline.from_pretrained(
        "SG161222/Realistic_Vision_V6.0_B1_noVAE",
        motion_adapter=motion_adapter,
        controlnet=controlnet,
        vae=vae,
        torch_dtype=torch.float16,
    ).to(device)
    
    pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config, beta_schedule="linear", algorithm_type="dpmsolver++", use_karras_sigmas=True)
    
    image_files = [
        "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatediff-scribble-1.png",
        "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatediff-scribble-2.png",
        "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatediff-scribble-3.png"
    ]
    conditioning_frames = [load_image(img_file) for img_file in image_files]

    conditioning_frame_indices = eval(conditioning_frame_indices)
    controlnet_conditioning_scale = float(controlnet_conditioning_scale)

    video = pipe(
        prompt=prompt,
        negative_prompt=negative_prompt,
        num_inference_steps=num_inference_steps,
        conditioning_frames=conditioning_frames,
        controlnet_conditioning_scale=controlnet_conditioning_scale,
        controlnet_frame_indices=conditioning_frame_indices,
        generator=torch.Generator().manual_seed(1337),
    ).frames[0]
    
    export_to_gif(video, "output.gif")
    return "output.gif"

demo = gr.Interface(
    fn=generate_video,
    inputs=[
        gr.Textbox(label="Prompt (한글 또는 영어)", value="카페에서 커피 마시는 아름다운 프랑스 여성, 걸작, 고품질"),
        gr.Textbox(label="Negative Prompt (한글 또는 영어)", value="저품질, 최악의 품질, 레터박스"),
        gr.Slider(label="Number of Inference Steps", minimum=1, maximum=200, step=1, value=100),
        gr.Textbox(label="Conditioning Frame Indices", value="[0, 8, 15]"),
        gr.Slider(label="ControlNet Conditioning Scale", minimum=0.1, maximum=2.0, step=0.1, value=1.0)
    ],
    outputs=gr.Image(label="Generated Video"),
    title="AnimateDiffSparseControlNetPipeline을 사용한 비디오 생성",
    description="AnimateDiffSparseControlNetPipeline을 사용하여 비디오를 생성합니다. 한글 또는 영어로 프롬프트를 입력할 수 있습니다."
)

demo.launch()