Update app.py

app.py

@@ -1,61 +1,101 @@
-import gradio as gr
-import subprocess
 import os
+import shutil
+import subprocess
 from pathlib import Path
+import gradio as gr
+
+REPO_DIR = Path("vrt")
+SETUP_FLAG = Path("setup_done.txt")
+OUTPUT_DIR = Path("outputs")
+OUTPUT_DIR.mkdir(exist_ok=True)
+
+def run(cmd, cwd=None):
+    """
+    Run a shell command safely and raise a helpful error if it fails.
+    """
+    p = subprocess.run(
+        cmd,
+        cwd=str(cwd) if cwd else None,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.STDOUT,
+        text=True,
+        check=False
+    )
+    if p.returncode != 0:
+        raise RuntimeError(f"Command failed ({p.returncode}): {' '.join(cmd)}\n\n{p.stdout}")
+    return p.stdout
 
 def setup():
-    if Path("setup_done.txt").exists():
+    """
+    One-time setup: clone repo + download models.
+    Uses a flag file so it doesn't redo work every request.
+    """
+    if SETUP_FLAG.exists() and REPO_DIR.exists():
         return
-    
+
     print("Setting up Video-Retalking...")
-    
-    # Clone repo
-    os.system("git clone https://github.com/OpenTalker/video-retalking.git vrt")
-    
-    # Download models
-    os.chdir("vrt")
-    os.system("bash scripts/download_models.sh")
-    os.chdir("..")
-    
-    Path("setup_done.txt").touch()
+
+    if not REPO_DIR.exists():
+        run(["git", "clone", "https://github.com/OpenTalker/video-retalking.git", str(REPO_DIR)])
+
+    # Download models (idempotent in most cases)
+    run(["bash", "scripts/download_models.sh"], cwd=REPO_DIR)
+
+    SETUP_FLAG.touch()
     print("✅ Setup complete!")
 
-def generate(image, audio):
-    if not image or not audio:
+def generate(image_path, audio_path):
+    if not image_path or not audio_path:
         return None, "❌ Upload both image and audio!"
-    
+
     try:
         setup()
-        
-        output = "result.mp4"
-        
+
+        image_path = Path(image_path).resolve()
+        audio_path = Path(audio_path).resolve()
+
+        if not image_path.exists():
+            return None, f"❌ Image not found: {image_path}"
+        if not audio_path.exists():
+            return None, f"❌ Audio not found: {audio_path}"
+
+        # Unique output name per run
+        out_path = (OUTPUT_DIR / "result.mp4").resolve()
+
+        # Clean old output if any
+        if out_path.exists():
+            out_path.unlink()
+
         # Run inference
-        cmd = f"cd vrt && python inference.py --face ../{image} --audio ../{audio} --outfile ../{output}"
-        
-        result = os.system(cmd)
-        
-        if Path(output).exists():
-            return output, "✅ Video generated successfully!"
-        else:
-            return None, "❌ Failed to generate video"
-            
+        cmd = [
+            "python",
+            "inference.py",
+            "--face", str(image_path),
+            "--audio", str(audio_path),
+            "--outfile", str(out_path),
+        ]
+        run(cmd, cwd=REPO_DIR)
+
+        if out_path.exists():
+            return str(out_path), "✅ Video generated successfully!"
+        return None, "❌ Failed to generate video (no output file created)."
+
     except Exception as e:
-        return None, f"❌ Error: {str(e)}"
+        return None, f"❌ Error: {e}"
 
-# Gradio Interface
 demo = gr.Interface(
     fn=generate,
     inputs=[
         gr.Image(type="filepath", label="📷 Face Image"),
-        gr.Audio(type="filepath", label="🎵 Audio File")
+        gr.Audio(type="filepath", label="🎵 Audio File"),
     ],
     outputs=[
         gr.Video(label="📹 Generated Video"),
-        gr.Textbox(label="Status", lines=2)
+        gr.Textbox(label="Status", lines=3),
     ],
     title="🎬 Video-Retalking Lip Sync",
-    description="Upload a face image and audio to generate lip-synced video"
+    description="Upload a face image and audio to generate a lip-synced video.",
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()