app.py

import gradio as gr
from transformers import pipeline
import requests
import json
import os

def speechToText(file):

    api_key = os.getenv("veni18sttts")  
    API_URL = "https://api-inference.huggingface.co/models/openai/whisper-large-v3-turbo"
    headers = {"Authorization": f"Bearer {api_key}"}

    def query(file):
        with open(file, "rb") as f:
            data = f.read()
        response = requests.post(API_URL, headers=headers, data=data)
        return response.json()

    my_text = query(file)

    #sentences = my_text["text"].split(".")

    
    def translate(sentences):
        translation = pipeline("translation", model="Helsinki-NLP/opus-mt-en-hu")
    
        text_translated=[]
        for text in sentences:
            text_translated.append(translation(text))

        return text_translated
    
    #text_translate = translate(sentences)

    #combined_text = ' '.join([item['translation_text'] for sublist in text_translated for item in sublist])

    def summa(my_text):
        pipe = pipeline("summarization", model="facebook/bart-large-cnn")

        max = 1024 
        if len(my_text) > max:    
            combined_text_parts = [my_text[i:i + max] for i in range(0, len(my_text), max)]
        else:
            combined_text_parts = [my_text]

        summaries = []
        for part in combined_text_parts:
            summary = pipe(part, max_length=180, min_length=140, do_sample=False)
            summaries.append(summary[0]['summary_text'])


        full_summary = ' '.join(summaries)


        return full_summary
    
    full_summaries = summa(my_text["text"])

    return full_summaries

demo = gr.Interface(fn=speechToText, inputs="file", outputs="text")
demo.launch()