app.py

import gradio as gr
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
from outetts.v0_1.interface import InterfaceHF, InterfaceGGUF


interface = InterfaceHF("OuteAI/OuteTTS-0.1-350M")
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large").to("cpu")


class Aspecto():
    pass

screen = Aspecto()
with gr.Blocks(theme=gr.themes.Ocean(primary_hue="pink", neutral_hue="indigo", font=[gr.themes.GoogleFont("Montserrat"), "Playwrite England SemiJoine", "Quicksand"])) as demo:
    textbox = gr.Textbox(label="Url")
    with gr.Row():
        button = gr.Button("Describir", variant="primary")
        clear = gr.Button("Borrar")
    output = gr.Textbox(label="Resumen")
    with gr.Row():
        button2 = gr.Button("Leer", variant="primary")
        clear = gr.Button("Borrar")
    output2 = gr.Audio(label="Audio")

    def describir(url):
      raw_image = Image.open(requests.get(url, stream=True).raw).convert('RGB')
      inputs = processor(raw_image, return_tensors="pt").to("cpu")
      out = model.generate(**inputs)
      return processor.decode(out[0], skip_special_tokens=True)

    def leer(texto):
        output = interface.generate(
        text= texto,
        temperature=0.1,
        repetition_penalty=1.1,
        max_length=4096
        )
        
        #output.play()
        #output.save("output.wav")
        
        return output


    button.click(describir, [textbox], output)
    button2.click(leer, [output], output2)

demo.launch(debug=True)