Update app.py

app.py

@@ -2,9 +2,12 @@ import gradio as gr
 import yt_dlp
 import os
 import shutil
+import subprocess
 from faster_whisper import WhisperModel
 
-# --- 1. Model Setup ---
+# ===============================
+# 1. Whisper Model (Lazy Load)
+# ===============================
 model = None
 
 def load_model():
@@ -12,116 +15,139 @@ def load_model():
     if model is None:
         print("📥 Loading Whisper Model...")
         model = WhisperModel("base", device="cpu", compute_type="int8")
-        print("✅ Model Loaded!")
+        print("✅ Model Loaded")
     return model
 
-# --- 2. Helper: Find FFmpeg ---
-def get_ffmpeg_dir():
-    # System me ffmpeg kahan hai, ye pata lagao
+# ===============================
+# 2. FFmpeg Path
+# ===============================
+def get_ffmpeg_path():
     path = shutil.which("ffmpeg")
-    if path:
-        return os.path.dirname(path) # Folder ka rasta return karo
-    return "/usr/bin" # Default fallback
-
-# --- 3. Logic: Download Audio from URL ---
+    return path if path else "/usr/bin/ffmpeg"
+
+# ===============================
+# 3. Convert Video → Audio
+# ===============================
+def extract_audio(video_path):
+    audio_path = "uploaded_audio.wav"
+    if os.path.exists(audio_path):
+        os.remove(audio_path)
+
+    cmd = [
+        get_ffmpeg_path(),
+        "-i", video_path,
+        "-vn",
+        "-ac", "1",
+        "-ar", "16000",
+        audio_path,
+        "-y"
+    ]
+    subprocess.run(cmd, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+    return audio_path
+
+# ===============================
+# 4. Download Audio from ANY URL
+# ===============================
 def download_audio_from_url(url):
-    output_path = "downloaded_audio"
-    if os.path.exists(f"{output_path}.mp3"): os.remove(f"{output_path}.mp3")
-
-    # Dynamic FFmpeg Path (Ye error fix karega)
-    ffmpeg_dir = get_ffmpeg_dir()
-    print(f"🔧 FFmpeg found at: {ffmpeg_dir}")
+    output = "url_audio.%(ext)s"
 
     ydl_opts = {
-        'format': 'bestaudio/best',
-        'outtmpl': output_path,
-        'ffmpeg_location': ffmpeg_dir, # <--- FIXED
-        'postprocessors': [{
-            'key': 'FFmpegExtractAudio',
-            'preferredcodec': 'mp3',
-            'preferredquality': '192',
+        "format": "bestaudio/best",
+        "outtmpl": output,
+        "ffmpeg_location": os.path.dirname(get_ffmpeg_path()),
+        "postprocessors": [{
+            "key": "FFmpegExtractAudio",
+            "preferredcodec": "wav",
+            "preferredquality": "192",
         }],
-        'quiet': True,
-        'no_warnings': True,
-        'nocheckcertificate': True,
-        'http_headers': {
-            'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36',
-            'Referer': 'https://www.tiktok.com/'
-        }
+        "quiet": True,
+        "nocheckcertificate": True,
     }
 
-    try:
-        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
-            ydl.download([url])
-        return f"{output_path}.mp3"
-    except Exception as e:
-        raise Exception(f"Download Fail: {str(e)}")
+    with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+        ydl.download([url])
 
-# --- 4. Main Transcribe Function (Handles Both) ---
+    return "url_audio.wav"
+
+# ===============================
+# 5. Main Transcribe Logic
+# ===============================
 def transcribe_media(url_input, file_input):
-    
-    audio_file_path = None
-    
+
     try:
-        # CASE 1: File Upload
-        if file_input is not None:
-            print(f"📂 Processing Uploaded File: {file_input}")
-            audio_file_path = file_input
-            
-        # CASE 2: URL Input
-        elif url_input and url_input.strip() != "":
-            print(f"🔗 Processing URL: {url_input}")
-            audio_file_path = download_audio_from_url(url_input)
-            
+        audio_path = None
+
+        # ---------- FILE UPLOAD ----------
+        if file_input:
+            ext = os.path.splitext(file_input)[1].lower()
+
+            if ext in [".mp3", ".wav", ".m4a"]:
+                audio_path = file_input
+            else:
+                audio_path = extract_audio(file_input)
+
+        # ---------- URL ----------
+        elif url_input and url_input.strip():
+            audio_path = download_audio_from_url(url_input)
+
         else:
-            return "⚠️ Error: Link daalein ya File upload karein."
-
-        if not os.path.exists(audio_file_path):
-             return "❌ Error: File nahi mili."
-
-        # --- Transcribe ---
-        current_model = load_model()
-        
-        # Turbo Settings: beam_size=1 (Fast)
-        segments, _ = current_model.transcribe(
-            audio_file_path, 
-            beam_size=1, 
+            return "⚠️ Please paste a link or upload a file."
+
+        if not os.path.exists(audio_path):
+            return "❌ Audio processing failed."
+
+        model = load_model()
+
+        segments, _ = model.transcribe(
+            audio_path,
+            beam_size=1,
             vad_filter=True
         )
-        
-        text = " ".join([s.text for s in segments])
-        return text.strip()
+
+        text = " ".join(seg.text for seg in segments)
+        return text.strip() if text else "⚠️ No speech detected."
 
     except Exception as e:
         return f"❌ Error: {str(e)}"
 
-# --- 5. UI ---
+# ===============================
+# 6. UI
+# ===============================
 css = """
 .container {max-width: 900px; margin: auto;}
-.gr-button-primary {background: linear-gradient(90deg, #1CB5E0 0%, #000851 100%); border: none; color: white;}
+.gr-button-primary {
+    background: linear-gradient(90deg,#ff416c,#ff4b2b);
+    border: none;
+    color: white;
+}
 """
 
 with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
-    
     with gr.Column(elem_classes="container"):
-        gr.Markdown("# 🚀 Turbo Transcriber (Link & Upload)")
-        gr.Markdown("Paste a TikTok link **OR** upload an Audio/Video file.")
+        gr.Markdown("## 🚀 Universal Video Transcript Tool")
+        gr.Markdown(
+            "Supports **YouTube, TikTok, Instagram, Facebook, Twitter/X**\n\n"
+            "**OR** upload video/audio file."
+        )
 
         with gr.Tabs():
-            # TAB 1: Link
             with gr.TabItem("🔗 Paste Link"):
-                url_in = gr.Textbox(label="TikTok / YouTube URL", placeholder="https://...")
-                btn_url = gr.Button("🚀 Transcribe Link", variant="primary")
+                url_in = gr.Textbox(
+                    label="Video URL",
+                    placeholder="https://youtube.com / tiktok.com / instagram.com"
+                )
+                btn_url = gr.Button("🎧 Transcribe Link", variant="primary")
 
-            # TAB 2: File Upload
             with gr.TabItem("📂 Upload File"):
-                file_in = gr.Audio(label="Upload File", type="filepath", sources=["upload", "microphone"])
+                file_in = gr.File(
+                    label="Upload Video / Audio",
+                    file_types=[".mp4", ".mkv", ".mov", ".webm", ".avi", ".mp3", ".wav"]
+                )
                 btn_file = gr.Button("📂 Transcribe File", variant="primary")
 
-        transcript_out = gr.Code(label="Transcript Result", language="markdown", interactive=False, lines=15)
+        output = gr.Code(label="Transcript Output", language="markdown", lines=15)
 
-    # Actions
-    btn_url.click(fn=transcribe_media, inputs=[url_in, gr.State(None)], outputs=transcript_out)
-    btn_file.click(fn=transcribe_media, inputs=[gr.State(None), file_in], outputs=transcript_out)
+    btn_url.click(transcribe_media, [url_in, gr.State(None)], output)
+    btn_file.click(transcribe_media, [gr.State(None), file_in], output)
 
-demo.launch()
+demo.launch()