import gradio as gr
import numpy as np
from huggingsound import SpeechRecognitionModel
from diffusers import StableDiffusionPipeline, DPMSolverMultistepScheduler
from transformers import pipeline
import librosa

# Función para convertir la tasa de muestreo del audio de entrada
def modelo1(audio):
    audio_data, sample_rate = audio
    # Asegurarse de que audio_data sea un array NumPy
    if not isinstance(audio_data, np.ndarray):
        audio_data = np.array(audio_data)

    # Convertir audio estéreo a mono
    if audio_data.shape[0] == 2:
        audio_data = np.mean(audio_data, axis=0)

    # Utilizar audio_data como entrada para el modelo
    whisper = pipeline('automatic-speech-recognition', model='openai/whisper-medium', device=-1)  # Cambia 'device' a -1 para usar la CPU
    text = whisper(audio_data, sample_rate)
    return text

def modelo2(text):
    model_id = "stabilityai/stable-diffusion-2-1"

    # Use the DPMSolverMultistepScheduler (DPM-Solver++) scheduler here instead
    pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
    pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
    pipe = pipe.to("cuda")

    image = pipe(text).images[0]
    return image

def execution(audio):
    modelo1res = modelo1(audio)
    modelo2res = modelo2(modelo1res)
    return modelo2res

if __name__ == "__main__":
    demo = gr.Interface(fn=modelo1, inputs="audio", outputs="text")
    demo.launch()