app.py

import gradio as gr
import torch
import random
from diffusers import DiffusionPipeline
from transformers import pipeline

device = "cuda" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
MAX_SEED = 2**32 - 1

# --- Model lists ---
image_models = {
    "Stable Diffusion 1.5": "runwayml/stable-diffusion-v1-5",
    "Stable Diffusion 2.1": "stabilityai/stable-diffusion-2-1",
    "SDXL Base 1.0": "stabilityai/stable-diffusion-xl-base-1.0",
    "Playground v2": "playgroundai/playground-v2-1024px-aesthetic",
    "Kandinsky 3": "kandinsky-community/kandinsky-3",
    "PixArt": "PixArt-alpha/PixArt-LCM-XL-2-1024-MS",
    "BLIP Diffusion": "Salesforce/blipdiffusion",
    "Muse 512": "amused/muse-512-finetuned",
    "Dreamlike 2.0": "dreamlike-art/dreamlike-photoreal-2.0",
    "OpenJourney": "prompthero/openjourney"
}

video_models = {
    "AnimateDiff": "animate-diff/animate-diff",
    "CogVideoX-5b": "THUDM/CogVideoX-5b",
    "HunyuanVideo": "tencent/HunyuanVideo",
    "LTX-Video": "Lightricks/LTX-Video",
    "ModelScope T2V": "damo-vilab/modelscope-text-to-video-synthesis",
    "VideoCrafter": "videocrafter/videocrafter",
    "Mochi-1": "mochi/mochi-1",
    "Allegro": "allegro/allegro",
    "OpenSora": "LanguageBind/Open-Sora-Plan-v1.2.0",
    "Zer0Scope": "zero-scope/zero-scope"
}

text_models = {
    "GPT-2": "gpt2",
    "GPT-Neo 1.3B": "EleutherAI/gpt-neo-1.3B",
    "GPT-J 6B": "EleutherAI/gpt-j-6B",
    "BLOOM 1.1B": "bigscience/bloom-1b1",
    "Falcon 7B": "tiiuae/falcon-7b",
    "MPT 7B": "mosaicml/mpt-7b",
    "LLaMA 2 7B": "meta-llama/Llama-2-7b-hf",
    "BTLM 3B": "cerebras/btlm-3b-8k-base",
    "XGen 7B": "Salesforce/xgen-7b-8k-base",
    "StableLM 2": "stabilityai/stablelm-2-1_6b"
}

# --- Caching loaded pipelines ---
image_pipes = {}
text_pipes = {}

# --- Functional logic ---
def generate_image(prompt, model_name, seed, randomize_seed):
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
    generator = torch.manual_seed(seed)

    if model_name not in image_pipes:
        image_pipes[model_name] = DiffusionPipeline.from_pretrained(
            image_models[model_name],
            torch_dtype=torch_dtype
        ).to(device)

    pipe = image_pipes[model_name]
    image = pipe(prompt=prompt, generator=generator, num_inference_steps=25, width=512, height=512).images[0]
    return image, seed

def generate_text(prompt, model_name):
    if model_name not in text_pipes:
        text_pipes[model_name] = pipeline("text-generation", model=text_models[model_name], device=0 if device == "cuda" else -1)
    pipe = text_pipes[model_name]
    output = pipe(prompt, max_length=100, do_sample=True)[0]['generated_text']
    return output

def generate_video(prompt, model_name):
    # Placeholder: real video models would return video frames
    return f"[Video placeholder] Model: {model_name}\nPrompt: {prompt}"

# --- Interface ---
with gr.Blocks() as demo:
    gr.Markdown("# 🔄 Multi-Task AI Generator")

    with gr.Tabs():
        # Tab 1: Image Generation
        with gr.Tab("🖼️ Image"):
            img_prompt = gr.Textbox(label="Prompt")
            img_model = gr.Dropdown(choices=list(image_models.keys()), value="Stable Diffusion 1.5", label="Select Image Model")
            img_seed = gr.Slider(0, MAX_SEED, value=42, label="Seed")
            img_rand = gr.Checkbox(label="Randomize seed", value=True)
            img_btn = gr.Button("Generate Image")
            img_out = gr.Image()
            img_btn.click(fn=generate_image, inputs=[img_prompt, img_model, img_seed, img_rand], outputs=[img_out, img_seed])

        # Tab 2: Video Generation
        with gr.Tab("🎥 Video"):
            vid_prompt = gr.Textbox(label="Prompt")
            vid_model = gr.Dropdown(choices=list(video_models.keys()), value="AnimateDiff", label="Select Video Model")
            vid_btn = gr.Button("Generate Video")
            vid_out = gr.Textbox(label="Result (Placeholder)")
            vid_btn.click(fn=generate_video, inputs=[vid_prompt, vid_model], outputs=vid_out)

        # Tab 3: Text Generation
        with gr.Tab("📝 Text"):
            txt_prompt = gr.Textbox(label="Prompt")
            txt_model = gr.Dropdown(choices=list(text_models.keys()), value="GPT-2", label="Select Text Model")
            txt_btn = gr.Button("Generate Text")
            txt_out = gr.Textbox(label="Generated Text")
            txt_btn.click(fn=generate_text, inputs=[txt_prompt, txt_model], outputs=txt_out)

demo.launch(show_error=True)