CoolFace
Apppublic

NoticIA-Col/Generador-Noticias

sourceHugging Facemitupdated 1y agoView on Hugging Face
1likes
App README

import os import openai import whisper import tempfile import gradio as gr from pydub import AudioSegment import fitz # PyMuPDF for handling PDFs import docx # For handling .docx files import pandas as pd # For handling .xlsx and .csv files import requests from bs4 import BeautifulSoup from moviepy.editor import VideoFileClip import yt_dlp import logging

Configure logging

logging.basicConfig(level=logging.INFO) logger = logging.getLogger(_name_)

Configure your OpenAI API key

openai.apikey = os.getenv("OPENAIAPI_KEY")

Load the highest quality Whisper model once

model = whisper.load_model("large")

def downloadsocialmediavideo(url): """Downloads a video from social media.""" ydlopts = { 'format': 'bestaudio/best', 'postprocessors': [{ 'key': 'FFmpegExtractAudio', 'preferredcodec': 'mp3', 'preferredquality': '192', }], 'outtmpl': '%(id)s.%(ext)s', } try: with ytdlp.YoutubeDL(ydlopts) as ydl: infodict = ydl.extractinfo(url, download=True) audiofile = f"{infodict['id']}.mp3" logger.info(f"Video successfully downloaded: {audiofile}") return audiofile except Exception as e: logger.error(f"Error downloading video: {str(e)}") raise

def convertvideotoaudio(videofile): """Converts a video file to audio.""" try: video = VideoFileClip(videofile) with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tempfile: video.audio.writeaudiofile(tempfile.name) logger.info(f"Video converted to audio: {tempfile.name}") return tempfile.name except Exception as e: logger.error(f"Error converting video to audio: {str(e)}") raise

def preprocessaudio(audiofile): """Preprocesses the audio file to improve quality.""" try: audio = AudioSegment.fromfile(audiofile) audio = audio.applygain(-audio.dBFS + (-20)) with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tempfile: audio.export(tempfile.name, format="mp3") logger.info(f"Audio preprocessed: {tempfile.name}") return temp_file.name except Exception as e: logger.error(f"Error preprocessing audio file: {str(e)}") raise

def transcribeaudio(file): """Transcribes an audio or video file.""" try: if isinstance(file, str) and file.startswith('http'): logger.info(f"Downloading social media video: {file}") filepath = downloadsocialmediavideo(file) elif isinstance(file, str) and file.lower().endswith(('.mp4', '.avi', '.mov', '.mkv')): logger.info(f"Converting local video to audio: {file}") filepath = convertvideotoaudio(file) else: logger.info(f"Preprocessing audio file: {file}") filepath = preprocess_audio(file)

logger.info(f"Transcribing audio: {filepath}") result = model.transcribe(filepath) transcription = result.get("text", "Error in transcription") logger.info(f"Transcription completed: {transcription[:50]}...") return transcription except Exception as e: logger.error(f"Error processing file: {str(e)}") return f"Error processing file: {str(e)}"

def readdocument(documentpath): """Reads content from PDF, DOCX, XLSX or CSV documents.""" try: if documentpath.endswith(".pdf"): doc = fitz.open(documentpath) return "\n".join([page.gettext() for page in doc]) elif documentpath.endswith(".docx"): doc = docx.Document(documentpath) return "\n".join([paragraph.text for paragraph in doc.paragraphs]) elif documentpath.endswith(".xlsx"): return pd.readexcel(documentpath).tostring() elif documentpath.endswith(".csv"): return pd.readcsv(documentpath).to_string() else: return "Unsupported file type. Please upload a PDF, DOCX, XLSX or CSV document." except Exception as e: return f"Error reading document: {str(e)}"

def readurl(url): """Reads content from a URL.""" try: response = requests.get(url) response.raiseforstatus() soup = BeautifulSoup(response.content, 'html.parser') return soup.gettext() except Exception as e: return f"Error reading URL: {str(e)}"

def processsocialcontent(url): """Processes content from a social media URL, handling both text and video.""" try: # First, try to read content as text textcontent = readurl(url)

# Then, try to process as video try: videocontent = transcribeaudio(url) except Exception: video_content = None

return { "text": textcontent, "video": videocontent } except Exception as e: logger.error(f"Error processing social content: {str(e)}") return None

def generatenews(instructions, facts, size, tone, *args): """Generates a news article from instructions, facts, URLs, documents, transcriptions, and social media content.""" knowledgebase = { "instructions": instructions, "facts": facts, "documentcontent": [], "audiodata": [], "urlcontent": [], "socialcontent": [] } numaudios = 5 * 3 # 5 audios/videos * 3 fields (file, name, position) numsocialurls = 3 * 3 # 3 social media URLs * 3 fields (URL, name, context) numurls = 5 # 5 general URLs audios = args[:numaudios] socialurls = args[numaudios:numaudios+numsocialurls] urls = args[numaudios+numsocialurls:numaudios+numsocialurls+numurls] documents = args[numaudios+numsocialurls+num_urls:]

for url in urls: if url: knowledgebase["urlcontent"].append(read_url(url))

for document in documents: if document is not None: knowledgebase["documentcontent"].append(read_document(document.name))

for i in range(0, len(audios), 3): audiofile, name, position = audios[i:i+3] if audiofile is not None: knowledgebase["audiodata"].append({"audio": audio_file, "name": name, "position": position})

for i in range(0, len(socialurls), 3): socialurl, socialname, socialcontext = socialurls[i:i+3] if socialurl: socialcontent = processsocialcontent(socialurl) if socialcontent: knowledgebase["socialcontent"].append({ "url": socialurl, "name": socialname, "context": socialcontext, "text": socialcontent["text"], "video": socialcontent["video"] }) logger.info(f"Social media content processed: {social_url}")

transcriptionstext, rawtranscriptions = "", ""

for idx, data in enumerate(knowledgebase["audiodata"]): if data["audio"] is not None: transcription = transcribeaudio(data["audio"]) transcriptiontext = f'"{transcription}" - {data["name"]}, {data["position"]}' rawtranscription = f'[Audio/Video {idx + 1}]: "{transcription}" - {data["name"]}, {data["position"]}' transcriptionstext += transcriptiontext + "\n" rawtranscriptions += raw_transcription + "\n\n"

for data in knowledgebase["socialcontent"]: if data["text"]: transcriptiontext = f'[Social media text]: "{data["text"][:200]}..." - {data["name"]}, {data["context"]}' transcriptionstext += transcriptiontext + "\n" rawtranscriptions += transcriptiontext + "\n\n" if data["video"]: transcriptionvideo = f'[Social media video]: "{data["video"]}" - {data["name"]}, {data["context"]}' transcriptionstext += transcriptionvideo + "\n" rawtranscriptions += transcriptionvideo + "\n\n"

documentcontent = "\n\n".join(knowledgebase["documentcontent"]) urlcontent = "\n\n".join(knowledgebase["urlcontent"])

internal_prompt = """ Instructions for the model:

  • —Follow news article principles: answer the 5 Ws in the first paragraph (Who?, What?, When?, Where?, Why?).
  • —Ensure at least 80% of quotes are direct and in quotation marks.
  • —The remaining 20% can be indirect quotes.
  • —Don't invent new information.
  • —Be rigorous with provided facts.
  • —When processing uploaded documents, extract and highlight important quotes and testimonials from sources.
  • —When processing uploaded documents, extract and highlight key figures.
  • —Avoid using the date at the beginning of the news body. Start directly with the 5Ws.
  • —Include social media content relevantly, citing the source and providing proper context.
  • —Make sure to relate the provided context for social media content with its corresponding transcription or text. """

prompt = f""" {internalprompt} Write a news article with the following information, including a title, a 15-word hook (additional information that complements the title), and the content body with {size} words. The tone should be {tone}. Instructions: {knowledgebase["instructions"]} Facts: {knowledgebase["facts"]} Additional content from documents: {documentcontent} Additional content from URLs: {urlcontent} Use the following transcriptions as direct and indirect quotes (without changing or inventing content): {transcriptionstext} """

try: response = openai.ChatCompletion.create( model="gpt-4o-mini", messages=[{"role": "user", "content": prompt}], temperature=0.1 ) news = response['choices'][0]['message']['content'] return news, raw_transcriptions except Exception as e: logger.error(f"Error generating news article: {str(e)}") return f"Error generating news article: {str(e)}", ""

with gr.Blocks() as demo: gr.Markdown("## All-in-One News Generator")

# Add tool description and attribution gr.Markdown(""" ### About this tool

This AI-powered news generator helps journalists and content creators produce news articles by processing multiple types of input:

  • —Audio and video files with automatic transcription
  • —Social media content
  • —Documents (PDF, DOCX, XLSX, CSV)
  • —Web URLs

The tool uses advanced AI to generate well-structured news articles following journalistic principles and maintaining the integrity of source quotes.

Created by Camilo Vega, AI Consultant """)

with gr.Row(): with gr.Column(scale=2): instructions = gr.Textbox(label="News article instructions", lines=2) facts = gr.Textbox(label="Describe the news facts", lines=4) size = gr.Number(label="Content body size (in words)", value=100) tone = gr.Dropdown(label="News tone", choices=["serious", "neutral", "lighthearted"], value="neutral") with gr.Column(scale=3): inputslist = [instructions, facts, size, tone] with gr.Tabs(): for i in range(1, 6): with gr.TabItem(f"Audio/Video {i}"): file = gr.File(label=f"Audio/Video {i}", type="filepath", filetypes=["audio", "video"]) name = gr.Textbox(label="Name", scale=1) position = gr.Textbox(label="Position", scale=1) inputslist.extend([file, name, position]) for i in range(1, 4): with gr.TabItem(f"Social Media {i}"): socialurl = gr.Textbox(label=f"Social media URL {i}", lines=1) socialname = gr.Textbox(label=f"Person/account name {i}", scale=1) socialcontext = gr.Textbox(label=f"Content context {i}", lines=2) inputslist.extend([socialurl, socialname, socialcontext]) for i in range(1, 6): with gr.TabItem(f"URL {i}"): url = gr.Textbox(label=f"URL {i}", lines=1) inputslist.append(url) for i in range(1, 6): with gr.TabItem(f"Document {i}"): document = gr.File(label=f"Document {i}", type="filepath", filecount="single") inputs_list.append(document)

gr.Markdown("---") # Visual separator

with gr.Row(): transcriptions_output = gr.Textbox(label="Transcriptions", lines=10)

gr.Markdown("---") # Visual separator

with gr.Row(): generate = gr.Button("Generate Draft") with gr.Row(): news_output = gr.Textbox(label="Generated Draft", lines=20)

generate.click(fn=generatenews, inputs=inputslist, outputs=[newsoutput, transcriptionsoutput])

Add description about how to use the app

gr.Markdown("""

How to Use This App

  1. 1.Input your requirements:
  2. 2.Enter your news article instructions
  3. 3.Describe the key facts of your news story
  4. 4.Set the desired word count and tone
  1. 1.Add your sources:
  2. 2.Upload audio/video files for automatic transcription
  3. 3.Add social media URLs to extract content
  4. 4.Include web URLs for additional information
  5. 5.Upload documents (PDF, DOCX, XLSX, CSV) to extract relevant data
  1. 1.Generate your draft:
  2. 2.Click "Generate Draft" to create your news article
  3. 3.Review the transcriptions to verify source accuracy
  4. 4.Use the generated draft as a starting point for your news story

This tool helps streamline the news writing process by automatically gathering, organizing, and synthesizing information from multiple sources into a cohesive article that follows journalistic best practices.

Created by Camilo Vega, AI Consultant """)

demo.launch(share=True)