app.py

from diffusers import AudioLDM2Pipeline
import torch
import gradio as gr

model = "cvssp/audioldm2"
pipe = AudioLDM2Pipeline.from_pretrained(model, torch_dtype = torch.float16,).to()

prompt = "A cheerful ukulele strumming in a beachside jam."

generator = torch.Generator().manual_seed(0)

#audio = pipe(prompt,audio_length_in_s = 10.24, generator = generator).audios[0]

def music_gen(prompt, duration=5):
    negative_prompt = "Low quality"
    audio = pipe(
        prompt,
        negative_prompt = negative_prompt,
        audio_length_in_s = duration, 
        generator = generator
    ).audios[0]
    return 16000, audio

interface = gr.Interface(
    examples = [["A cheerful ukulele strumming in a beachside jam."],["A violin playing a heartfelt melody."]],
    fn = music_gen,
    inputs = [gr.Textbox(),gr.Slider(1, 120, value=5, step=1, label="Duration")],
    outputs = "audio",
).launch(debug=True, share=True)