commit

app.py

@@ -455,9 +455,6 @@ h1,h2,h3,h4,h5,h6 {
 footer {
     display: none !important;
 }
-#col-container {
-    margin: 0 auto;
-}
 .image-container {
     aspect-ratio: """,str(width),"/",str(height),""" !important;
 }
@@ -522,7 +519,7 @@ def upscaler(
     tile_width: int = 112,
     tile_height: int = 144,
     denoise_strength: float = 0.35,
-    num_inference_steps: int = 30,
+    num_inference_steps: int = 18,
     solver: str = "DDIM",
 ) -> Image.Image:
 
@@ -545,7 +542,7 @@ def upscaler(
         tile_size=(tile_height, tile_width),
         denoise_strength=denoise_strength,
         num_inference_steps=num_inference_steps,
-        loras_scale={"more_details": 1.0, "sdxl_render": 1.0},
+        loras_scale={"more_details": 0.5, "sdxl_render": 1.0},
         solver_type=solver_type,
     )
 
@@ -585,42 +582,42 @@ def pipe_generate_image(p1,p2):
     log(f'RET pipe_generate')
     return imgs
 
-def add_song_cover_text(img,artist,song):
+def add_song_cover_text(img,artist,song,height,width):
 
     draw = ImageDraw.Draw(img)
 
     rows = 1
-    labels_distance = 1 / 2.5
+    labels_distance = 1/3
 
     textheight=min(math.ceil( width / 10 ), math.ceil( height / 5 ))
     font = ImageFont.truetype(r"Alef-Bold.ttf", textheight)
     textwidth = draw.textlength(song,font)
     x = math.ceil((width - textwidth) / 2)
-    y = height - math.ceil(textheight * rows / 2)
-    y = y - math.ceil(y * labels_distance)
+    y = height - (textheight * rows / 2) - (height / 2)
+    y = math.ceil(y - (height / 2 * labels_distance))
     draw.text((x, y), song, (255,255,255), font=font, spacing=2, stroke_width=4, stroke_fill=(0,0,0))
 
     textheight=min(math.ceil( width / 12 ), math.ceil( height / 6 ))
     font = ImageFont.truetype(r"Alef-Bold.ttf", textheight)
     textwidth = draw.textlength(artist,font)
     x = math.ceil((width - textwidth) / 2)
-    y = height - math.ceil(textheight * rows / 2)
-    y = y + math.ceil(y * labels_distance)
+    y = height - (textheight * rows / 2) - (height / 2)
+    y = math.ceil(y + (height / 2 * labels_distance))
     draw.text((x, y), artist, (0,0,0), font=font, spacing=4, stroke_width=2, stroke_fill=(255,255,255))
 
     return img
 
-@spaces.GPU(duration=240)
+@spaces.GPU(duration=180)
 def all_pipes(pos,neg,artist,song):
     imgs = pipe_generate_image(pos,neg)
 
     names = []
     index = 1
     for img in imgs:
-        labeled_img = add_song_cover_text(img,artist,song)
-        enhanced_img = upscaler(labeled_img)
+        enhanced_img = upscaler(img)
+        labeled_img = add_song_cover_text(enhanced_img,artist,song,height*4,width*4)
         name = f'{artist} - {song} ({index}).png'
-        enhanced_img.save(name)
+        labeled_img.save(name)
         names.append(name)
     return names
 
@@ -636,7 +633,7 @@ def handle_generate(artist,song,genre,lyrics):
     pos_lyrics_sum = pos_lyrics if pos_lyrics == "" else summarize_text(pos_lyrics)
     neg = f"Textual Labeled Distorted Discontinuous Ugly Blurry Low-Quality Worst-Quality Low-Resolution Painted"
     q = "\""
-    pos = f'Realistic Vivid Genuine Reasonable Highly-Detailed 4K { pos_genre } SONG { pos_song } { "INSTRUMENTAL" if pos_lyrics_sum == "" else q + pos_lyrics_sum + q }'
+    pos = f'Realistic Vivid Genuine Reasonable Highly-Detailed 4K { pos_genre } { pos_song }{ pos_lyrics_sum if pos_lyrics_sum == "" else " " + q + pos_lyrics_sum + q }'
 
     print(f"""
         Positive: {pos}
@@ -653,21 +650,20 @@ if __name__ == "__main__":
         gr.Markdown(f"""
             # Song Cover Image Generator
         """)
-        with gr.Column():
-            with gr.Column():
-                with gr.Row():
-                    artist = gr.Textbox(
+        with gr.Column(scale=2):
+            artist = gr.Textbox(
                         placeholder="Artist name",
                         value="",
                         container=False,
                         max_lines=1
-                    )
-                    song = gr.Textbox(
+            )
+            song = gr.Textbox(
                         placeholder="Song name",
                         value="",
                         container=False,
                         max_lines=1
-                    )
+            )
+        with gr.Column(scale=2):
                 genre = gr.Textbox(
                         placeholder="Genre (English)",
                         value="",
@@ -681,10 +677,10 @@ if __name__ == "__main__":
                     max_lines=1
                 )
 
-            run = gr.Button("Generate",elem_classes="btn")
+        with gr.Column():
+            cover = gr.Image(interactive=False,container=False,elem_classes="image-container", label="Result", show_label=True, type='filepath', show_share_button=False)
 
-            with gr.Row():
-                cover = gr.Image(interactive=False,container=False,elem_classes="image-container", label="Result", show_label=True, type='filepath', show_share_button=False)
+        run = gr.Button("Generate",elem_classes="btn")
 
         run.click(
             fn=handle_generate,