app.py

from diffusers import AudioLDMPipeline
import torch
import gradio as gr
from googletrans import Translator
import os
os.environ["TOKENIZERS_PARALLELISM"] = "false"


if torch.cuda.is_available():
    device = "cuda"
    torch_dtype = torch.float16
else:
    device = "cpu"
    torch_dtype = torch.float32
print(device)
repo_id = "cvssp/audioldm-m-full"
pipe = AudioLDMPipeline.from_pretrained(repo_id, torch_dtype=torch_dtype)
pipe = pipe.to(device)
# pipe.unet = torch.compile(pipe.unet)
#pipe.unet = torch.compile(pipe.unet)

def generate_sound(text): 
    print(text)
    # text=translate_text(text)
    translator = Translator()
    text=translator.translate(text, src='es',dest="en").text
    print(text)
    waveforms = pipe(text, 
                     num_inference_steps=10, 
                     audio_length_in_s=5,
                     negative_prompt = "low quality, average quality").audios
    rate =16000
    return rate, waveforms[0]
    #return gr.make_waveform((rate, waveforms[0]))

def translate_text(text):
    translator = Translator()
    translated_text=translator.translate(text, src='es',dest="en")
    return translated_text.text

demo = gr.Blocks()
with demo:
    with gr.Row():
        with gr.Column():

            text = gr.Textbox(value="Ingrese el texto:")
            button = gr.Button(value="Generar")

        with gr.Column():
            output = gr.Audio()
            #output = gr.Video(label="Output")
    button.click(generate_sound,text,output)


demo.launch()