app.py

import gradio as gr
import torch
from diffusers import StableDiffusionPipeline
from pydub import AudioSegment

# === Use GPU if available ===
device = "cuda" if torch.cuda.is_available() else "cpu"

# === Load model ===
model_id = "stabilityai/stable-diffusion-2-1"
pipe = StableDiffusionPipeline.from_pretrained(model_id)
pipe.to(device)

# === Define function ===
def generate_image(audio, prompt):
    if audio is None:
        return None

    # Save audio temporarily
    audio_path = "train.wav"
    audio.export(audio_path, format="wav")

   
    result = pipe(prompt, guidance_scale=7.5, num_inference_steps=30).images[0]
    return result


interface = gr.Interface(
    fn=generate_image,
    inputs=[
        gr.Audio(source="upload", type="pydub"),
        gr.Textbox(label="Prompt", value="A surreal dreamscape made of music"),
    ],
    outputs=gr.Image(type="pil"),
    title="🎧 SonicDiffusion: Audio → Image Generator"
)

interface.launch()