Update app.py

app.py

@@ -290,16 +290,16 @@ def preset_changed(preset):
     ]    
 
 def generate_video_from_text(
-    prompt="",
-    enhance_prompt_toggle=False,
-    negative_prompt="low quality, worst quality, deformed, distorted, warped, motion smear, motion artifacts, fused fingers, incorrect anatomy, strange hands, unattractive",
-    frame_rate=25,
-    seed=171198,
-    num_inference_steps=41,
-    guidance_scale=4,
-    height=320,
-    width=512,
-    num_frames=257,
+    prompt,
+    enhance_prompt_toggle,
+    negative_prompt,
+    frame_rate,
+    seed,
+    num_inference_steps,
+    guidance_scale,
+    height,
+    width,
+    num_frames,
     progress=gr.Progress(),
 ):
     if len(prompt.strip()) < 50:
@@ -308,10 +308,23 @@ def generate_video_from_text(
             duration=5,
         )
 
+    # 프롬프트 개선이 활성화된 경우
+    if enhance_prompt_toggle:
+        prompt = enhance_prompt(prompt, "t2v")
+
     # Translate Korean prompts to English
     prompt = translate_korean_prompt(prompt)
     negative_prompt = translate_korean_prompt(negative_prompt)
 
+    # 기본값 설정
+    height = height or 320
+    width = width or 512
+    num_frames = num_frames or 257
+    frame_rate = frame_rate or 25
+    seed = seed or 171198
+    num_inference_steps = num_inference_steps or 41
+    guidance_scale = guidance_scale or 4.0
+
     sample = {
         "prompt": prompt,
         "prompt_attention_mask": None,
@@ -354,7 +367,6 @@ def generate_video_from_text(
         gc.collect()
 
     output_path = tempfile.mktemp(suffix=".mp4")
-    print(images.shape)
     video_np = images.squeeze(0).permute(1, 2, 3, 0).cpu().float().numpy()
     video_np = (video_np * 255).astype(np.uint8)
     height, width = video_np.shape[1:3]
@@ -371,21 +383,20 @@ def generate_video_from_text(
 
 def generate_video_from_image(
     image_path,
-    prompt="",
-    enhance_prompt_toggle=False,
-    negative_prompt="low quality, worst quality, deformed, distorted, warped, motion smear, motion artifacts, fused fingers, incorrect anatomy, strange hands, unattractive",
-    frame_rate=25,
-    seed=171198,
-    num_inference_steps=41,
-    guidance_scale=4,
-    height=320,
-    width=512,
-    num_frames=257,
+    prompt,
+    enhance_prompt_toggle,
+    negative_prompt,
+    frame_rate,
+    seed,
+    num_inference_steps,
+    guidance_scale,
+    height,
+    width,
+    num_frames,
     progress=gr.Progress(),
 ):
-    print("Height: ", height)
-    print("Width: ", width)
-    print("Num Frames: ", num_frames)
+    if not image_path:
+        raise gr.Error("입력 이미지를 제공해주세요.", duration=5)
 
     if len(prompt.strip()) < 50:
         raise gr.Error(
@@ -393,13 +404,24 @@ def generate_video_from_image(
             duration=5,
         )
 
-    if not image_path:
-        raise gr.Error("입력 이미지를 제공해주세요.", duration=5)
+    # 프롬프트 개선이 활성화된 경우
+    if enhance_prompt_toggle:
+        prompt = enhance_prompt(prompt, "i2v")
 
     # Translate Korean prompts to English
     prompt = translate_korean_prompt(prompt)
     negative_prompt = translate_korean_prompt(negative_prompt)
 
+    # 기본값 설정
+    height = height or 320
+    width = width or 512
+    num_frames = num_frames or 257
+    frame_rate = frame_rate or 25
+    seed = seed or 171198
+    num_inference_steps = num_inference_steps or 41
+    guidance_scale = guidance_scale or 4.0
+
+    # 이미지 로드 및 전처리
     media_items = (
         load_image_to_tensor_with_resize(image_path, height, width).to(device).detach()
     )
@@ -447,6 +469,7 @@ def generate_video_from_image(
         for frame in video_np[..., ::-1]:
             out.write(frame)
         out.release()
+
     except Exception as e:
         raise gr.Error(
             f"비디오 생성 중 오류가 발생했습니다. 다시 시도해주세요. 오류: {e}",
@@ -456,6 +479,12 @@ def generate_video_from_image(
     finally:
         torch.cuda.empty_cache()
         gc.collect()
+        if 'images' in locals():
+            del images
+        if 'video_np' in locals():
+            del video_np
+        if 'media_items' in locals():
+            del media_items
 
     return output_path
 
@@ -813,7 +842,7 @@ with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange") as iface:
                         lines=5,
                     )
                     txt2vid_enhance_toggle = Toggle(
-                        label="프롬프트 개선",
+                        label="프롬프트 증강",
                         value=False,
                         interactive=True,
                     )
@@ -1025,35 +1054,46 @@ with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange") as iface:
         outputs=txt2vid_prompt
     )
 
+# Event handlers 부분 수정
     txt2vid_generate.click(
         fn=generate_video_from_text,
         inputs=[
-            txt2vid_prompt,
-            txt2vid_enhance_toggle,
-            txt2vid_negative_prompt,
-            txt2vid_frame_rate,
-            *txt2vid_advanced[:3],
-            txt2vid_current_height,
-            txt2vid_current_width,
-            txt2vid_current_num_frames,
+            txt2vid_prompt,  # 텍스트 입력
+            txt2vid_enhance_toggle,  # 프롬프트 개선 토글
+            txt2vid_negative_prompt,  # 네거티브 프롬프트
+            txt2vid_frame_rate,  # 프레임 레이트
+            txt2vid_advanced[0],  # seed
+            txt2vid_advanced[1],  # inference_steps
+            txt2vid_advanced[2],  # guidance_scale
+            height_slider,  # height
+            width_slider,  # width
+            num_frames_slider,  # num_frames
         ],
         outputs=txt2vid_output,
-        concurrency_limit=1,
-        concurrency_id="generate_video",
-        queue=True,
+        api_name="generate_text_to_video"
     )
 
-    # Image to Video Tab handlers
-    img2vid_preset.change(
-        fn=preset_changed,
-        inputs=[img2vid_preset],
-        outputs=[
-            img2vid_current_height,
-            img2vid_current_width,
-            img2vid_current_num_frames,
-            *img2vid_advanced[3:]
-        ]
+# Image to Video 이벤트 핸들러 수정
+    img2vid_generate.click(
+        fn=generate_video_from_image,
+        inputs=[
+            img2vid_image,  # 입력 이미지
+            img2vid_prompt,  # 텍스트 입력
+            img2vid_enhance_toggle,  # 프롬프트 개선 토글
+            img2vid_negative_prompt,  # 네거티브 프롬프트
+            img2vid_frame_rate,  # 프레임 레이트
+            img2vid_advanced[0],  # seed
+            img2vid_advanced[1],  # inference_steps
+            img2vid_advanced[2],  # guidance_scale
+            height_slider,  # height
+            width_slider,  # width
+            num_frames_slider,  # num_frames
+        ],
+        outputs=img2vid_output,
+        api_name="generate_image_to_video"
     )
+    
+
 
     img2vid_enhance_toggle.change(
         fn=update_prompt_i2v,