Update app.py

app.py

@@ -11,19 +11,141 @@ print("PYDANTIC =", pydantic.__version__)
 print("HF HUB =", huggingface_hub.__version__)
 
 
-
 import gradio as gr
+import subprocess
+import os
+from PIL import Image
+import numpy as np
+from pydub import AudioSegment
+
+
+# ----------------------------
+# Save audio (numpy -> mp3)
+# ----------------------------
+def save_audio_mp3(audio_tuple, filename):
+    sampling_rate, audio_data = audio_tuple
+
+    audio_bytes = np.array(audio_data, dtype=np.int16).tobytes()
+
+    audio_segment = AudioSegment(
+        audio_bytes,
+        sample_width=2,
+        frame_rate=sampling_rate,
+        channels=1
+    )
+
+    audio_segment.export(filename, format="mp3")
+
+
+# ----------------------------
+# Merge video + audio correctly
+# ----------------------------
+def merge_audio_video(video_path, audio_path, output_path):
+
+    if os.path.exists(output_path):
+        os.remove(output_path)
+
+    cmd = [
+        "ffmpeg",
+        "-y",
+        "-i", video_path,
+        "-i", audio_path,
+        "-c:v", "copy",
+        "-c:a", "aac",
+        "-map", "0:v:0",
+        "-map", "1:a:0",
+        output_path
+    ]
+
+    subprocess.run(cmd, check=True)
+
+    return output_path
+
+
+# ----------------------------
+# Main inference
+# ----------------------------
+def run_inference(input_image, input_audio):
+
+    if input_image is None:
+        raise gr.Error("Please upload an image.")
+
+    if input_audio is None:
+        raise gr.Error("Please upload audio.")
+
+    os.makedirs("sample_data", exist_ok=True)
+    os.makedirs("results", exist_ok=True)
+
+    # Save image
+    image_path = "sample_data/uploaded_image.png"
+    Image.fromarray(input_image.astype(np.uint8)).save(image_path)
+
+    # Save audio
+    audio_path = "sample_data/uploaded_audio.mp3"
+    save_audio_mp3(input_audio, audio_path)
+
+    # Run Wav2Lip inference
+    cmd = [
+        "python3",
+        "inference.py",
+        "--checkpoint_path", "checkpoints/wav2lip_gan.pth",
+        "--face", image_path,
+        "--audio", audio_path
+    ]
+
+    result = subprocess.run(cmd, capture_output=True, text=True)
+
+    if result.returncode != 0:
+        raise gr.Error(f"Inference failed:\n{result.stderr}")
+
+    # output from wav2lip
+    wav2lip_video = "results/result_voice.mp4"
+
+    if not os.path.exists(wav2lip_video):
+        raise gr.Error("Wav2Lip output video not found!")
+
+    # merge audio + video
+    final_video = merge_audio_video(
+        wav2lip_video,
+        audio_path,
+        "results/final_output.mp4"
+    )
+
+    return final_video
+
+
+# ----------------------------
+# UI
+# ----------------------------
+def create_demo():
+
+    with gr.Blocks() as demo:
+
+        gr.Markdown("# 🎤 Wav2Lip Demo")
+
+        with gr.Row():
+            input_image = gr.Image(type="numpy", label="Image")
+            input_audio = gr.Audio(type="numpy", label="Audio")
+            output_video = gr.Video(label="Output", type="filepath")
+
+        btn = gr.Button("Generate")
+
+        btn.click(
+            fn=run_inference,
+            inputs=[input_image, input_audio],
+            outputs=output_video
+        )
+
+        gr.Markdown("### Sample")
+        with gr.Row():
+            gr.Image("sample/spark.png")
+            gr.Audio("sample/spark_1.1.mp3")
+            gr.Video("sample/final_output.mp4")
 
-def test(img, audio):
-    return "test.mp4"
+    return demo
 
-demo = gr.Interface(
-    fn=test,
-    inputs=[
-        gr.Image(type="numpy"),
-        gr.Audio(type="numpy")
-    ],
-    outputs=gr.Video()
-)
 
-demo.launch(show_api=False)
+if __name__ == "__main__":
+    demo = create_demo()
+    demo.queue()
+    demo.launch(show_api=False)