Update app.py

app.py

@@ -1,4 +1,4 @@
-# app.py — DeepFake AI Forensics (Light Royal Blue Theme)
+# app.py — DeepFake AI Forensics (WhatsApp Voice Note Fix)
 import os
 import subprocess
 import tempfile
@@ -30,24 +30,55 @@ model.to(device)
 print(f"[+] Model loaded on {device}")
 
 # ==========================================
-# 2. AUDIO / VIDEO PREPROCESSING
+# 2. AUDIO / VIDEO PREPROCESSING (ROBUST)
 # ==========================================
+def normalize_audio(file_path):
+    """
+    Converts ANY audio/video to standard 16kHz mono WAV via FFmpeg.
+    This fixes WhatsApp voice notes (Opus/OGG disguised as MP3), 
+    corrupt headers, and exotic codecs.
+    """
+    out = tempfile.mktemp(suffix=".wav")
+    cmd = [
+        "ffmpeg", "-y",
+        "-i", file_path,
+        "-vn",                      # no video
+        "-acodec", "pcm_s16le",     # 16-bit PCM
+        "-ar", "16000",             # 16 kHz
+        "-ac", "1",                 # mono
+        "-af", "loudnorm=I=-16:TP=-1.5:LRA=11",  # normalize levels
+        out
+    ]
+    result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    
+    if result.returncode != 0:
+        err = result.stderr.decode('utf-8', errors='ignore')[:200]
+        raise RuntimeError(f"FFmpeg could not decode this file. It may be corrupted or use an unsupported codec.\nDetails: {err}")
+    
+    return out
+
 def convert_to_audio(file_path):
     ext = os.path.splitext(file_path)[1].lower().lstrip('.')
-    if ext in ["wav", "mp3", "flac", "m4a", "ogg", "aac", "wma"]:
-        return file_path
-    if ext in ["mp4", "mkv", "avi", "mov", "webm", "flv"]:
-        print("[+] Video detected → extracting audio via ffmpeg...")
-        out = tempfile.mktemp(suffix=".wav")
-        cmd = [
-            "ffmpeg", "-i", file_path,
-            "-vn", "-acodec", "pcm_s16le",
-            "-ar", "16000", "-ac", "1",
-            out, "-y"
-        ]
-        subprocess.run(cmd, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, check=True)
-        return out
-    raise ValueError(f"Unsupported file format: {ext}")
+    
+    # Supported audio formats (including WhatsApp Opus)
+    audio_exts = ["wav", "mp3", "flac", "m4a", "ogg", "opus", "aac", "wma", "oga"]
+    # Supported video formats
+    video_exts = ["mp4", "mkv", "avi", "mov", "webm", "flv", "3gp"]
+    
+    if ext in audio_exts:
+        print(f"[+] Audio detected ({ext}) → normalizing via ffmpeg...")
+        return normalize_audio(file_path)
+    
+    if ext in video_exts:
+        print("[+] Video detected → extracting & normalizing audio via ffmpeg...")
+        return normalize_audio(file_path)
+    
+    # Unknown extension? Try ffmpeg anyway as last resort
+    print(f"[+] Unknown format ({ext}) → attempting ffmpeg decode...")
+    try:
+        return normalize_audio(file_path)
+    except Exception:
+        raise ValueError(f"Unsupported file format: {ext}. Please upload MP3, WAV, M4A, OGG, OPUS, or MP4.")
 
 def load_audio(path):
     audio, _ = librosa.load(path, sr=16000)
@@ -82,19 +113,19 @@ def analyze(file_path):
     if ai_score < 0.35:
         verdict = "HUMAN VOICE"
         level = "LOW RISK"
-        color = "#059669"       # emerald-600
+        color = "#059669"
         icon = "🧑"
         glow = "rgba(5,150,105,0.18)"
     elif ai_score < 0.65:
         verdict = "UNCERTAIN / MIXED"
         level = "MEDIUM RISK"
-        color = "#d97706"       # amber-600
+        color = "#d97706"
         icon = "⚠️"
         glow = "rgba(217,119,6,0.18)"
     else:
         verdict = "AI / SYNTHETIC VOICE"
         level = "HIGH RISK"
-        color = "#dc2626"       # red-600
+        color = "#dc2626"
         icon = "🤖"
         glow = "rgba(220,38,38,0.18)"
 
@@ -262,7 +293,12 @@ def detect_audio(audio_file):
         return plot_path, result_html, f"{percentage:.1f}%", color, ""
 
     except Exception as e:
-        return None, f"<div style='color:#dc2626;padding:30px;'>❌ Analysis Error: {str(e)}</div>", "Error", "#dc2626", ""
+        err_msg = str(e)
+        if "FFmpeg" in err_msg:
+            err_html = f"<div style='color:#dc2626;padding:30px;'><strong>❌ File Decode Error</strong><br><br>{err_msg}<br><br><span style='color:#475569;font-size:0.9em;'>WhatsApp voice notes are often .opus or .ogg files disguised as .mp3. Try renaming the file to .ogg or exporting it differently.</span></div>"
+        else:
+            err_html = f"<div style='color:#dc2626;padding:30px;'>❌ Analysis Error: {err_msg}</div>"
+        return None, err_html, "Error", "#dc2626", ""
 
 
 def detect_video(video_file):
@@ -347,7 +383,12 @@ def detect_video(video_file):
         return plot_path, result_html, f"{percentage:.1f}%", color, ""
 
     except Exception as e:
-        return None, f"<div style='color:#dc2626;padding:30px;'>❌ Analysis Error: {str(e)}</div>", "Error", "#dc2626", ""
+        err_msg = str(e)
+        if "FFmpeg" in err_msg:
+            err_html = f"<div style='color:#dc2626;padding:30px;'><strong>❌ File Decode Error</strong><br><br>{err_msg}</div>"
+        else:
+            err_html = f"<div style='color:#dc2626;padding:30px;'>❌ Analysis Error: {err_msg}</div>"
+        return None, err_html, "Error", "#dc2626", ""
 
 
 # ==========================================
@@ -497,11 +538,15 @@ def build_ui():
                             <span class="format-badge">M4A</span>
                             <span class="format-badge">FLAC</span>
                             <span class="format-badge">OGG</span>
+                            <span class="format-badge">OPUS</span>
                         </div>
                         <div style="font-size: 0.8em; color: #475569; margin-bottom: 16px; display: flex; align-items: center; gap: 6px;">
                             <span style="font-size: 1.2em;">📎</span>
                             <span>Maximum file size: <strong style="color: #1e40af;">50 MB</strong></span>
                         </div>
+                        <div style="font-size: 0.75em; color: #94a3b8; background: #eff6ff; border-radius: 8px; padding: 10px 12px; margin-bottom: 12px; line-height: 1.5;">
+                            💡 <strong>WhatsApp voice notes:</strong> If your file fails to upload, try renaming it from <code>.mp3</code> to <code>.ogg</code> or <code>.opus</code> before uploading.
+                        </div>
                         """)
                         
                         audio_input = gr.Audio(