Update app.py

app.py

@@ -50,40 +50,6 @@ system_prompt_i2v = """당신은 이미지 기반 비디오 생성을 위한 프
 촬영 감독이 촬영 목록을 설명하는 것처럼 구체적이고 시각적으로 작성하세요.
 200단어를 넘지 않도록 하되, 최대한 상세하게 작성하세요."""
 
-# Default preset
-default_preset = "16:9 (512x320)"
-
-# preset_options 수정 - 각 해상도별 정확한 프레임 수 적용
-preset_options = [
-    # 16:9 비율 (최대/최소)
-    {"label": "16:9 HD (1216x684)", "width": 1216, "height": 684, "num_frames": 41, "aspect": "16:9"},  # 1.6초
-    {"label": "16:9 (512x320)", "width": 512, "height": 320, "num_frames": 257, "aspect": "16:9"},  # 10.3초
-    
-    # 4:3 비율 (최대/최소)
-    {"label": "4:3 (1024x768)", "width": 1024, "height": 768, "num_frames": 49, "aspect": "4:3"},  # 2.0초
-    {"label": "4:3 (640x480)", "width": 640, "height": 480, "num_frames": 121, "aspect": "4:3"},  # 4.8초
-    
-    # 1:1 비율 (최대/최소)
-    {"label": "1:1 (896x896)", "width": 896, "height": 896, "num_frames": 73, "aspect": "1:1"},  # 2.9초
-    {"label": "1:1 (512x512)", "width": 512, "height": 512, "num_frames": 233, "aspect": "1:1"},  # 9.3초
-    
-    # 3:2 비율 (최대)
-    {"label": "3:2 (1200x800)", "width": 1200, "height": 800, "num_frames": 41, "aspect": "3:2"},  # 1.6초
-    
-    # 9:16 비율 (최소)
-    {"label": "9:16 (432x768)", "width": 432, "height": 768, "num_frames": 241, "aspect": "9:16"}  # 9.6초
-]
-
-# State 변수들의 초기값도 수정 (512x320 기준)
-txt2vid_current_height = gr.State(value=320)
-txt2vid_current_width = gr.State(value=512)
-txt2vid_current_num_frames = gr.State(value=257)  # 10.3초
-
-img2vid_current_height = gr.State(value=320)
-img2vid_current_width = gr.State(value=512)
-img2vid_current_num_frames = gr.State(value=257)  # 10.3초
-
-
 # Load Hugging Face token if needed
 hf_token = os.getenv("HF_TOKEN")
 openai_api_key = os.getenv("OPENAI_API_KEY")
@@ -229,17 +195,68 @@ pipeline = XoraVideoPipeline(
     vae=vae,
 ).to(device)
 
+
+# txt2vid와 img2vid의 현재 상태 값들을 수정
+txt2vid_current_height = gr.State(value=320)  # 수정됨
+txt2vid_current_width = gr.State(value=512)   # 수정됨
+txt2vid_current_num_frames = gr.State(value=257)  # 10.3초
+
+img2vid_current_height = gr.State(value=320)  # 수정됨
+img2vid_current_width = gr.State(value=512)   # 수정됨
+img2vid_current_num_frames = gr.State(value=257)  # 10.3초
+
+# Preset options for resolution and frame configuration
+# Convert frames to seconds assuming 25 FPS
+preset_options = [
+    {"label": "[16:9 HD] 1216x704, 1.6초", "width": 1216, "height": 704, "num_frames": 41},
+    {"label": "[16:9] 1088x704, 2.0초", "width": 1088, "height": 704, "num_frames": 49},
+    {"label": "[16:9] 1056x640, 2.3초", "width": 1056, "height": 640, "num_frames": 57},
+    {"label": "[16:9] 992x608, 2.6초", "width": 992, "height": 608, "num_frames": 65},
+    {"label": "[16:9] 896x608, 2.9초", "width": 896, "height": 608, "num_frames": 73},
+    {"label": "[16:9] 896x544, 3.2초", "width": 896, "height": 544, "num_frames": 81},
+    {"label": "[16:9] 832x544, 3.6초", "width": 832, "height": 544, "num_frames": 89},
+    {"label": "[16:9] 800x512, 3.9초", "width": 800, "height": 512, "num_frames": 97},
+    {"label": "[16:9] 768x512, 3.9초", "width": 768, "height": 512, "num_frames": 97},
+    {"label": "[16:9] 800x480, 4.2초", "width": 800, "height": 480, "num_frames": 105},
+    {"label": "[16:9] 736x480, 4.5초", "width": 736, "height": 480, "num_frames": 113},
+    {"label": "[3:2] 704x480, 4.8초", "width": 704, "height": 480, "num_frames": 121},
+    {"label": "[16:9] 704x448, 5.2초", "width": 704, "height": 448, "num_frames": 129},
+    {"label": "[16:9] 672x448, 5.5초", "width": 672, "height": 448, "num_frames": 137},
+    {"label": "[16:9] 640x416, 6.1초", "width": 640, "height": 416, "num_frames": 153},
+    {"label": "[16:9] 672x384, 6.4초", "width": 672, "height": 384, "num_frames": 161},
+    {"label": "[16:9] 640x384, 6.8초", "width": 640, "height": 384, "num_frames": 169},
+    {"label": "[16:9] 608x384, 7.1초", "width": 608, "height": 384, "num_frames": 177},
+    {"label": "[16:9] 576x384, 7.4초", "width": 576, "height": 384, "num_frames": 185},
+    {"label": "[16:9] 608x352, 7.7초", "width": 608, "height": 352, "num_frames": 193},
+    {"label": "[16:9] 576x352, 8.0초", "width": 576, "height": 352, "num_frames": 201},
+    {"label": "[16:9] 544x352, 8.4초", "width": 544, "height": 352, "num_frames": 209},
+    {"label": "[3:2] 512x352, 9.3초", "width": 512, "height": 352, "num_frames": 233},
+    {"label": "[16:9] 544x320, 9.6초", "width": 544, "height": 320, "num_frames": 241},
+    {"label": "[16:9] 512x320, 10.3초", "width": 512, "height": 320, "num_frames": 257},
+]
+
 def preset_changed(preset):
-    selected = next(item for item in preset_options if item["label"] == preset)
-    return (
-        selected["height"],
-        selected["width"],
-        selected["num_frames"],
-        gr.update(visible=False),
-        gr.update(visible=False),
-        gr.update(visible=False),
-    )
-    
+    if preset != "Custom":
+        selected = next(item for item in preset_options if item["label"] == preset)
+        # height, width, num_frames 값을 global 변수로 업데이트
+        return (
+            selected["height"],
+            selected["width"],
+            selected["num_frames"],
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+        )
+    else:
+        return (
+            None,
+            None,
+            None,
+            gr.update(visible=True),
+            gr.update(visible=True),
+            gr.update(visible=True),
+        )
+
 
 def generate_video_from_text(
     prompt="",
@@ -249,12 +266,11 @@ def generate_video_from_text(
     seed=171198,
     num_inference_steps=41,
     guidance_scale=4,
-    height=320,      # 기본값 수정
-    width=512,       # 기본값 수정
-    num_frames=257,  # 기본값 수정 (10.3초)
+    height=512,
+    width=320,
+    num_frames=257,
     progress=gr.Progress(),
 ):
-
     if len(prompt.strip()) < 50:
         raise gr.Error(
             "프롬프트는 최소 50자 이상이어야 합니다. 더 자세한 설명을 제공해주세요.",
@@ -295,7 +311,8 @@ def generate_video_from_text(
                 vae_per_channel_normalize=True,
                 conditioning_method=ConditioningMethod.UNCONDITIONAL,
                 mixed_precision=True,
-                callback_on_step_end=gradio_progress_callback,).images
+                callback_on_step_end=gradio_progress_callback,
+            ).images
     except Exception as e:
         raise gr.Error(
             f"비디오 생성 중 오류가 발생했습니다. 다시 시도해주세요. 오류: {e}",
@@ -330,12 +347,11 @@ def generate_video_from_image(
     seed=171198,
     num_inference_steps=50,
     guidance_scale=4,
-    height=320,      # 기본값 수정
-    width=512,       # 기본값 수정
-    num_frames=257,  # 기본값 수정 (10.3초)
+    height=512,
+    width=768,
+    num_frames=121,
     progress=gr.Progress(),
 ):
-
     print("Height: ", height)
     print("Width: ", width)
     print("Num Frames: ", num_frames)
@@ -439,26 +455,26 @@ def create_advanced_options():
         )
         height_slider = gr.Slider(
             label="4.4 Height",
-            minimum=320,
-            maximum=896,
+            minimum=256,
+            maximum=1024,
             step=64,
-            value=320,
+            value=512,
             visible=False,
         )
         width_slider = gr.Slider(
             label="4.5 Width",
-            minimum=512,
-            maximum=1216,
+            minimum=256,
+            maximum=1024,
             step=64,
-            value=512,
+            value=768,
             visible=False,
         )
         num_frames_slider = gr.Slider(
             label="4.5 Number of Frames",
-            minimum=41,      # 최소 프레임 수 (HD 해상도 기준)
-            maximum=257,     # 최대 프레임 수 (512x320 해상도 기준)
+            minimum=1,
+            maximum=200,
             step=1,
-            value=257,      # 기본값을 최대 프레임으로 설정 (10.3초)
+            value=121,
             visible=False,
         )
 
@@ -471,7 +487,6 @@ def create_advanced_options():
             num_frames_slider,
         ]
 
-
 # Gradio Interface Definition
 with gr.Blocks(theme=gr.themes.Soft()) as iface:
     with gr.Tabs():
@@ -486,7 +501,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as iface:
                         lines=5,
                     )
                     txt2vid_enhance_toggle = Toggle(
-                        label="프롬프트 증강",
+                        label="프롬프트 개선",
                         value=False,
                         interactive=True,
                     )
@@ -500,13 +515,13 @@ with gr.Blocks(theme=gr.themes.Soft()) as iface:
                     )
 
                     # 현재 선택된 값들을 저장할 상태 변수들
-                    txt2vid_current_height = gr.State(value=360)
-                    txt2vid_current_width = gr.State(value=640)
-                    txt2vid_current_num_frames = gr.State(value=81)
+                    txt2vid_current_height = gr.State(value=512)
+                    txt2vid_current_width = gr.State(value=320)
+                    txt2vid_current_num_frames = gr.State(value=257)
 
                     txt2vid_preset = gr.Dropdown(
                         choices=[p["label"] for p in preset_options],
-                        value=default_preset,
+                        value="512x320, 10.3초",
                         label="Step 2: 해상도 프리셋 선택",
                     )
 
@@ -558,13 +573,13 @@ with gr.Blocks(theme=gr.themes.Soft()) as iface:
                     )
 
                     # 현재 선택된 값들을 저장할 상태 변수들
-                    img2vid_current_height = gr.State(value=360)
-                    img2vid_current_width = gr.State(value=640)
-                    img2vid_current_num_frames = gr.State(value=81)
+                    img2vid_current_height = gr.State(value=512)
+                    img2vid_current_width = gr.State(value=768)
+                    img2vid_current_num_frames = gr.State(value=97)
 
                     img2vid_preset = gr.Dropdown(
                         choices=[p["label"] for p in preset_options],
-                        value=default_preset,
+                        value="512x320, 10.3초",
                         label="Step 3: 해상도 프리셋 선택",
                     )
 
@@ -662,4 +677,4 @@ with gr.Blocks(theme=gr.themes.Soft()) as iface:
 if __name__ == "__main__":
     iface.queue(max_size=64, default_concurrency_limit=1, api_open=False).launch(
         share=True, show_api=False
-    )                
+    )