Update app.py

app.py

@@ -2,15 +2,32 @@ import gradio as gr
 import whisper
 import yt_dlp
 import os
+import re
+import requests
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 from transformers import pipeline
 
-# Load Whisper model
+# Load models
 whisper_model = whisper.load_model("base")
+multilingual_model = "csebuetnlp/mT5_multilingual_XLSum"
+tokenizer = AutoTokenizer.from_pretrained(multilingual_model)
+summarizer_model = AutoModelForSeq2SeqLM.from_pretrained(multilingual_model)
+translator = pipeline("translation", model="Helsinki-NLP/opus-mt-mul-en")
 
-# Load Summarization pipeline (you can replace with a LLaMA fine-tuned model if you have one)
-summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
+SUPPORTED_LANGUAGES = {
+    "bn": "Bengali",
+    "en": "English",
+    "gu": "Gujarati",
+    "hi": "Hindi",
+    "kn": "Kannada",
+    "ml": "Malayalam",
+    "mr": "Marathi",
+    "ta": "Tamil",
+    "te": "Telugu",
+    "ur": "Urdu"
+}
 
-def download_youtube_audio(youtube_url):
+def download_audio(youtube_url):
     filename = "audio.mp3"
     ydl_opts = {
         'format': 'bestaudio/best',
@@ -22,39 +39,75 @@ def download_youtube_audio(youtube_url):
             'preferredquality': '192',
         }],
     }
-
     with yt_dlp.YoutubeDL(ydl_opts) as ydl:
         ydl.download([youtube_url])
-
     return filename
 
-def transcribe_and_summarize(youtube_url):
-    try:
-        # Step 1: Download audio
-        audio_file = download_youtube_audio(youtube_url)
+def extract_thumbnail(youtube_url):
+    video_id = None
+    match = re.search(r"(?:v=|\/)([0-9A-Za-z_-]{11}).*", youtube_url)
+    if match:
+        video_id = match.group(1)
+        return f"https://img.youtube.com/vi/{video_id}/0.jpg"
+    return ""
+
+def summarize_text(text, lang_code):
+    input_text = f"summarize: {text}"
+    inputs = tokenizer.encode(input_text, return_tensors="pt", max_length=512, truncation=True)
+    summary_ids = summarizer_model.generate(inputs, max_length=150, min_length=30, num_beams=4, early_stopping=True)
+    return tokenizer.decode(summary_ids[0], skip_special_tokens=True)
 
-        # Step 2: Transcribe
+def transcribe_and_summarize(youtube_url, translate_to_english):
+    try:
+        audio_file = download_audio(youtube_url)
         result = whisper_model.transcribe(audio_file)
         transcript = result["text"]
+        lang_code = result["language"]
+
+        thumbnail_url = extract_thumbnail(youtube_url)
+
+        if lang_code not in SUPPORTED_LANGUAGES:
+            return None, f"❌ Language '{lang_code}' not supported.", "", ""
 
-        # Step 3: Summarize
-        summary = summarizer(transcript, max_length=200, min_length=50, do_sample=False)[0]["summary_text"]
+        summary = summarize_text(transcript, lang_code)
+
+        if translate_to_english and lang_code != "en":
+            translated_summary = translator(summary)[0]["translation_text"]
+        else:
+            translated_summary = summary
 
-        # Clean up
         os.remove(audio_file)
 
-        return f"📄 **Transcript:**\n\n{transcript}\n\n🧠 **Summary:**\n\n{summary}"
+        summary_text = f"Transcript:\n{transcript}\n\nSummary:\n{translated_summary}"
+
+        # Save to file
+        with open("summary.txt", "w", encoding="utf-8") as f:
+            f.write(summary_text)
+
+        return thumbnail_url, f"🗣️ Language: {SUPPORTED_LANGUAGES[lang_code]}", transcript, translated_summary, "summary.txt"
 
     except Exception as e:
-        return f"❌ Error: {str(e)}"
+        return None, f"❌ Error: {str(e)}", "", "", None
 
 # Gradio UI
-demo = gr.Interface(
-    fn=transcribe_and_summarize,
-    inputs=gr.Textbox(label="Enter YouTube Video URL"),
-    outputs=gr.Markdown(label="Transcript and Summary"),
-    title="🎥 YouTube Video Summarizer",
-    description="Enter a YouTube video link. This app will transcribe it and summarize using AI."
-)
+with gr.Blocks(css="style.css") as demo:
+    gr.Markdown("<h1 style='text-align: center;'>🎬 Multilingual YouTube Summarizer</h1>")
+    gr.Markdown("This tool transcribes any YouTube video and summarizes it in its original language or English.")
+
+    with gr.Row():
+        youtube_url = gr.Textbox(label="YouTube Video URL")
+        translate_check = gr.Checkbox(label="Translate Summary to English", value=True)
+
+    thumbnail = gr.Image(label="Video Thumbnail", type="filepath")
+    lang_out = gr.Text(label="Detected Language")
+    transcript_out = gr.Textbox(label="Transcript", lines=8)
+    summary_out = gr.Textbox(label="Summary", lines=6)
+    download_btn = gr.File(label="Download .txt")
+
+    btn = gr.Button("Generate Summary")
+
+    btn.click(fn=transcribe_and_summarize,
+              inputs=[youtube_url, translate_check],
+              outputs=[thumbnail, lang_out, transcript_out, summary_out, download_btn])
 
 demo.launch()