Update app.py

app.py

@@ -9,18 +9,19 @@ from PIL import Image
 import torch
 
 # Initialize the model
-cmd = 'python3 -m pip install "huggingface_hub[cli]"; \
-huggingface-cli download LeonJoe13/Sonic --local-dir  checkpoints; \
-huggingface-cli download stabilityai/stable-video-diffusion-img2vid-xt --local-dir  checkpoints/stable-video-diffusion-img2vid-xt; \
-huggingface-cli download openai/whisper-tiny --local-dir checkpoints/whisper-tiny;'
+cmd = (
+    'python3 -m pip install "huggingface_hub[cli]"; '
+    'huggingface-cli download LeonJoe13/Sonic --local-dir  checkpoints; '
+    'huggingface-cli download stabilityai/stable-video-diffusion-img2vid-xt --local-dir  checkpoints/stable-video-diffusion-img2vid-xt; '
+    'huggingface-cli download openai/whisper-tiny --local-dir checkpoints/whisper-tiny;'
+)
 os.system(cmd)
 
 pipe = Sonic()
 
 def get_md5(content):
     md5hash = hashlib.md5(content)
-    md5 = md5hash.hexdigest()
-    return md5
+    return md5hash.hexdigest()
 
 @spaces.GPU(duration=300)  # Increased duration to handle longer videos
 def get_video_res(img_path, audio_path, res_video_path, dynamic_scale=1.0):
@@ -28,9 +29,9 @@ def get_video_res(img_path, audio_path, res_video_path, dynamic_scale=1.0):
     min_resolution = 512
     inference_steps = 25
 
-    # Get audio duration
+    # Get audio duration (정보 출력용)
     audio = AudioSegment.from_file(audio_path)
-    duration = len(audio) / 1000.0  # Convert to seconds
+    duration = len(audio) / 1000.0  # 초 단위 변환
 
     face_info = pipe.preprocess(img_path, expand_ratio=expand_ratio)
     print(f"Face detection info: {face_info}")
@@ -42,15 +43,14 @@ def get_video_res(img_path, audio_path, res_video_path, dynamic_scale=1.0):
         img_path = crop_image_path
         os.makedirs(os.path.dirname(res_video_path), exist_ok=True)
 
-        # Process with full audio duration
+        # NOTE: Sonic.process()는 더 이상 duration 인자를 받지 않으므로 제거합니다.
         pipe.process(
-            img_path,
-            audio_path,
-            res_video_path,
+            img_path, 
+            audio_path, 
+            res_video_path, 
             min_resolution=min_resolution,
             inference_steps=inference_steps,
-            dynamic_scale=dynamic_scale,
-            duration=duration  # Pass the actual duration
+            dynamic_scale=dynamic_scale
         )
     else:
         return -1
@@ -61,21 +61,21 @@ os.makedirs(tmp_path, exist_ok=True)
 os.makedirs(res_path, exist_ok=True)
 
 def process_sonic(image, audio, dynamic_scale):
-    # Input validation
+    # 입력 검증
     if image is None:
         raise gr.Error("Please upload an image")
     if audio is None:
         raise gr.Error("Please upload an audio file")
-
+        
     img_md5 = get_md5(np.array(image))
     audio_md5 = get_md5(audio[1])
     print(f"Processing with image hash: {img_md5}, audio hash: {audio_md5}")
-
+    
     sampling_rate, arr = audio[:2]
     if len(arr.shape) == 1:
         arr = arr[:, None]
-
-    # Create audio segment
+    
+    # 오디오 세그먼트 생성
     audio_segment = AudioSegment(
         arr.tobytes(),
         frame_rate=sampling_rate,
@@ -83,19 +83,19 @@ def process_sonic(image, audio, dynamic_scale):
         channels=arr.shape[1]
     )
     audio_segment = audio_segment.set_frame_rate(sampling_rate)
-
-    # Generate paths
+    
+    # 파일 경로 생성
     image_path = os.path.abspath(os.path.join(tmp_path, f'{img_md5}.png'))
     audio_path = os.path.abspath(os.path.join(tmp_path, f'{audio_md5}.wav'))
     res_video_path = os.path.abspath(os.path.join(res_path, f'{img_md5}_{audio_md5}_{dynamic_scale}.mp4'))
-
-    # Save inputs if they don't exist
+    
+    # 입력 파일이 없으면 저장
     if not os.path.exists(image_path):
         image.save(image_path)
     if not os.path.exists(audio_path):
         audio_segment.export(audio_path, format="wav")
-
-    # Process or return cached result
+    
+    # 캐시된 결과가 있으면 반환, 없으면 새로 생성
     if os.path.exists(res_video_path):
         print(f"Using cached result: {res_video_path}")
         return res_video_path
@@ -103,12 +103,10 @@ def process_sonic(image, audio, dynamic_scale):
         print(f"Generating new video with dynamic scale: {dynamic_scale}")
         return get_video_res(image_path, audio_path, res_video_path, dynamic_scale)
 
-# Dummy get_example function to prevent errors if examples are not defined
+# 예시 데이터를 위한 dummy 함수 (필요에 따라 실제 예시 데이터를 넣으세요)
 def get_example():
-    # 예시가 없다면 빈 리스트를 반환하거나 실제 예시 데이터를 입력할 수 있습니다.
     return []
 
-# Enhanced UI
 css = """
 .gradio-container {
     font-family: 'Arial', sans-serif;
@@ -136,7 +134,7 @@ with gr.Blocks(css=css) as demo:
             <p>Transform still images into dynamic videos synchronized with audio</p>
         </div>
     """)
-
+    
     with gr.Row():
         with gr.Column():
             image_input = gr.Image(
@@ -144,13 +142,13 @@ with gr.Blocks(css=css) as demo:
                 label="Portrait Image",
                 elem_id="image_input"
             )
-
+            
             audio_input = gr.Audio(
                 label="Voice/Audio Input",
                 elem_id="audio_input",
                 type="numpy"
             )
-
+            
             with gr.Column():
                 dynamic_scale = gr.Slider(
                     minimum=0.5,
@@ -160,28 +158,28 @@ with gr.Blocks(css=css) as demo:
                     label="Animation Intensity",
                     info="Adjust to control movement intensity (0.5: subtle, 2.0: dramatic)"
                 )
-
+            
             process_btn = gr.Button(
                 "Generate Animation",
                 variant="primary",
                 elem_id="process_btn"
             )
-
+        
         with gr.Column():
             video_output = gr.Video(
                 label="Generated Animation",
                 elem_id="video_output"
             )
-
-    # Process button click
+    
+    # 버튼 클릭 시 애니메이션 생성 함수 호출
     process_btn.click(
         fn=process_sonic,
         inputs=[image_input, audio_input, dynamic_scale],
         outputs=video_output,
         api_name="animate"
     )
-
-    # Examples section (elem_classes 인자 제거)
+    
+    # 예시 섹션
     gr.Examples(
         examples=get_example(),
         fn=process_sonic,
@@ -189,8 +187,8 @@ with gr.Blocks(css=css) as demo:
         outputs=video_output,
         cache_examples=False
     )
-
-    # Footer with attribution and links
+    
+    # Footer: Attribution & Links
     gr.HTML("""
         <div style="text-align: center; margin-top: 2em;">
             <div style="margin-bottom: 1em;">
@@ -205,4 +203,5 @@ with gr.Blocks(css=css) as demo:
         </div>
     """)
 
-demo.launch()
+# 공유 링크 생성: share=True
+demo.launch(share=True)