app.py

import os
from huggingface_hub import login
import torch
import torchaudio
from einops import rearrange
import gradio as gr
from stable_audio_tools import get_pretrained_model
from stable_audio_tools.inference.generation import generate_diffusion_cond

# Authenticate using HF token from secrets
login(token=os.getenv("HUGGINGFACE_TOKEN"))

device = "cuda" if torch.cuda.is_available() else "cpu"
model, config = get_pretrained_model("stabilityai/stable-audio-open-small")
model = model.to(device)
sample_rate = config["sample_rate"]
sample_size = config["sample_size"]

def generate_audio(prompt):
    conditioning = [{"prompt": prompt, "seconds_total": 11}]
    with torch.no_grad():
        output = generate_diffusion_cond(
            model,
            steps=8,
            conditioning=conditioning,
            sample_size=sample_size,
            device=device
        )
    output = rearrange(output, "b d n -> d (b n)")
    output = output.to(torch.float32).div(torch.max(torch.abs(output))).clamp(-1, 1).mul(32767).to(torch.int16).cpu()
    path = "output.wav"
    torchaudio.save(path, output, sample_rate)
    return path

gr.Interface(
    fn=generate_audio,
    inputs=gr.Textbox(label="Enter your sound prompt"),
    outputs=gr.Audio(type="filepath"),
    title="Stable Audio Generator"
).launch()