app.py

from dotenv import find_dotenv, load_dotenv
from transformers import pipeline
from langchain import PromptTemplate, LLMChain, OpenAI
import requests
import os
import streamlit as st
import json


load_dotenv (find_dotenv())
HUGGINGFACEHUB_API_TOKEN = os.getenv("HUGGINGFACEHUB_API_TOKEN")

#img2text

def img2text(url):
    image_to_text = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")

    text = image_to_text(url)[0]["generated_text"]

    print(text)
    return text

#llm

def generate_story(scenario):
    template = """
    Vous êtes un conteur ;
    Vous pouvez créer une histoire courte à partir d'un récit simple. L'histoire ne doit pas dépasser 50 mots;

    CONTEXT: {scenario}
    STORY:
    """

    prompt = PromptTemplate(template=template, input_variables=["scenario"])

    story_llm = LLMChain(llm=OpenAI(
        model_name="gpt-3.5-turbo", temperature=1), prompt=prompt, verbose=True)
    
    story = story_llm.predict(scenario=scenario)

    print(story)
    return story

	

#text to speech

def text2speech(message):
     #API_URL = "https://api-inference.huggingface.co/models/espnet/kan-bayashi_ljspeech_vits"
     API_URL = "https://api-inference.huggingface.co/models/facebook/mms-tts-fra"
     headers = {"Authorization": f"Bearer {HUGGINGFACEHUB_API_TOKEN}"}
     payloads = {
     "inputs": message
     }

     response = requests.post(API_URL, headers=headers, json=payloads)
     with open('audio.wav', 'wb') as file:
         file.write(response.content)

#translate
def translatefr(Text2img2text):
     API_URL = "https://api-inference.huggingface.co/models/sgugger/marian-finetuned-kde4-en-to-fr"
     headers = {"Authorization": f"Bearer {HUGGINGFACEHUB_API_TOKEN}"}
     payloads = {
     "inputs": Text2img2text
     }

     response = requests.post(API_URL, headers=headers, json=payloads)
     responsejson = response.content
     # Votre donnée JSON encodée en bytes
     data_bytes = responsejson

     # Décoder les bytes en string
     data_str = data_bytes.decode('utf-8')

     # Analyser la chaîne JSON pour obtenir un objet Python
     data = json.loads(data_str)

     # Extraire le texte souhaité
     text = data[0]['translation_text']
     print(text)
     return text

#scenario = img2text("mmd.png")
#story = generate_story(scenario)
#en_fr_translator = pipeline("translation_en_to_fr")
#story_fr = en_fr_translator(story)[0]["translation_text"]
#print(story_fr)
#text2speech(story_fr)



def main():

    st.set_page_config(page_title="Img 2 audio story")

    st.header("Turn img into audio story")
    uploaded_file = st.file_uploader("Choose an image....", type="jpg")

    if uploaded_file is not None:
        print(uploaded_file)
        bytes_data = uploaded_file.getvalue()
        with open(uploaded_file.name, "wb") as file:
            file.write(bytes_data)
        st.image(uploaded_file, caption='Uploaded Image.',
                 use_column_width=True)
        scenario = img2text(uploaded_file.name)
        scenariofr = translatefr(scenario)
        story = generate_story(scenariofr)
        #en_fr_translator = pipeline("translation_en_to_fr")
        #story_fr = en_fr_translator(story)[0]["translation_text"]
        text2speech(story)

        with st.expander("scenario"):
            st.write(scenariofr)
        with st.expander("story"):
            st.write(story)
        
        st.audio("audio.wav")

if __name__ == '__main__':
    main()