app.py

import spaces
import os
import tempfile
import gradio as gr
from dotenv import load_dotenv
import torch
from scipy.io.wavfile import write
from diffusers import DiffusionPipeline
from transformers import pipeline
from pathlib import Path

load_dotenv()
hf_token = os.getenv("HF_TKN")

device_id = 0 if torch.cuda.is_available() else -1

captioning_pipeline = pipeline(
    "image-to-text",
    model="nlpconnect/vit-gpt2-image-captioning",
    device=device_id
)

pipe = DiffusionPipeline.from_pretrained(
    "cvssp/audioldm2",
    use_auth_token=hf_token
)

@spaces.GPU(duration=120)
def analyze_image_with_free_model(image_file):
    try:
        with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as temp_file:
            temp_file.write(image_file)
            temp_image_path = temp_file.name

        results = captioning_pipeline(temp_image_path)
        if not results or not isinstance(results, list):
            return "Error: Could not generate caption.", True
        
        caption = results[0].get("generated_text", "").strip()
        if not caption:
            return "No caption was generated.", True
        return caption, False

    except Exception as e:
        return f"Error analyzing image: {e}", True

@spaces.GPU(duration=120)
def get_audioldm_from_caption(caption):
    try:
        pipe.to("cuda")
        audio_output = pipe(
            prompt=caption,
            num_inference_steps=50,
            guidance_scale=7.5
        )
        pipe.to("cpu")
        audio = audio_output.audios[0]

        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_wav:
            write(temp_wav.name, 16000, audio)
            return temp_wav.name

    except Exception as e:
        print(f"Error generating audio from caption: {e}")
        return None

css = """
#header-container {
    text-align: center;
    margin: 20px 0;
}

#header-title {
    font-size: 36px;
    font-weight: bold;
    margin-bottom: 10px;
    color: #333333;
}

#header-subtitle {
    font-size: 18px;
    margin-bottom: 20px;
    color: #555555;
}

#main-container {
    max-width: 900px;
    margin: 0 auto;
    padding: 20px;
    border-radius: 12px;
    background: linear-gradient(135deg, #ffffff, #f0f0f0);
    box-shadow: 0 4px 8px rgba(0, 0, 0, 0.2);
}

button.primary-button {
    background: linear-gradient(90deg, #007bff, #0056b3);
    color: white;
    border: none;
    padding: 12px 24px;
    border-radius: 8px;
    font-size: 16px;
    cursor: pointer;
    font-weight: bold;
    transition: transform 0.2s, box-shadow 0.2s;
}

button.primary-button:hover {
    transform: translateY(-3px);
    box-shadow: 0 8px 15px rgba(0, 0, 0, 0.2);
}

button.primary-button:active {
    transform: translateY(0);
    box-shadow: none;
}

#footer-container {
    margin-top: 30px;
    text-align: center;
    color: #666666;
    font-size: 14px;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="header-container"):
        gr.HTML("""
        <div id="header-title">🎶 Image-to-Sound Generator</div>
        <div id="header-subtitle">Transform your images into descriptive captions and immersive soundscapes.</div>
        """)

    with gr.Box(elem_id="main-container"):
        gr.Markdown("""
        ### How It Works
        1. **Upload an Image**: Select an image to analyze.
        2. **Generate Description**: Get a detailed caption describing your image.
        3. **Generate Sound**: Create an audio representation based on the caption.
        """)

        image_upload = gr.File(label="Upload Image", type="binary")
        generate_description_button = gr.Button("Generate Description", elem_classes="primary-button")
        caption_display = gr.Textbox(label="Generated Caption", interactive=False, placeholder="Your image caption will appear here.")
        generate_sound_button = gr.Button("Generate Sound", elem_classes="primary-button")
        audio_output = gr.Audio(label="Generated Sound Effect", interactive=False)

    with gr.Box(elem_id="footer-container"):
        gr.Markdown("""
        ## About This App
        This application uses advanced machine learning models to transform images into text captions and generate matching sound effects. It's a unique blend of visual and auditory creativity, powered by state-of-the-art AI technology.


        For inquiries, contact us at [[email protected]](mailto:[email protected]).
        """)

    def update_caption(image_file):
        description, _ = analyze_image_with_free_model(image_file)
        return description

    def generate_sound(description):
        if not description or description.startswith("Error"):
            return None
        audio_path = get_audioldm_from_caption(description)
        return audio_path

    generate_description_button.click(
        fn=update_caption,
        inputs=image_upload,
        outputs=caption_display
    )

    generate_sound_button.click(
        fn=generate_sound,
        inputs=caption_display,
        outputs=audio_output
    )

demo.launch(debug=True, share=True)