format lyric input

app.py

@@ -63,18 +63,32 @@ def generate_song(lyric, description=None, prompt_audio=None, genre=None, cfg_co
     params = {k:v for k,v in params.items() if v is not None}
     sample_rate = MODEL.cfg.sample_rate
     
-    # 适配lyric格式
-    lyric = re.sub(r"[^\w\s\[\]\-\u4e00-\u9fff\u3040-\u309f\u30a0-\u30ff\uac00-\ud7af\u00c0-\u017f]", "", lyric)
-    lyric = lyric.lower()
+    # format lyric
     lyric = lyric.replace("[intro]", "[intro-short]").replace("[inst]", "[inst-short]").replace("[outro]", "[outro-short]")
-    lyric = lyric.replace("\n\n", " ; ")
-    for s in STRUCTS:
-        lyric = lyric.replace(f"{s}\n", f"{s} ")
-    lyric = lyric.replace("\n", ".")
-    lyric = lyric.replace(". ; ", " ; ")
-    lyric = lyric.replace("].", "] ").replace("..", " ")
-    
-    # 适配prompt 
+    paragraphs = [p.strip() for p in lyric.strip().split('\n\n') if p.strip()]
+    paragraphs_norm = []
+    for para in paragraphs:
+        lines = para.splitlines()
+        struct_tag = lines[0].strip().lower()
+        if struct_tag not in STRUCTS:
+            return None, json.dumps(f"segments should start with a structure tag in {STRUCTS}")
+        if struct_tag in ['[verse]', '[chorus]', '[bridge]']:
+            if len(lines) < 2 or not [line.strip() for line in lines[1:] if line.strip()]:
+                return None, json.dumps("The following segments require lyrics: [verse], [chorus], [bridge]")
+            else:
+                new_para_list = []
+                for line in lines[1:]:
+                    new_para_list.append(re.sub(r"[^\w\s\[\]\-\u4e00-\u9fff\u3040-\u309f\u30a0-\u30ff\uac00-\ud7af\u00c0-\u017f]", "", line))
+                new_para_str = f"{struct_tag} {'.'.join(new_para_list)}"
+        else:
+            if len(lines) > 1:
+                return None, json.dumps("The following segments should not contain lyrics: [intro], [intro-short], [intro-medium], [inst], [inst-short], [inst-medium], [outro], [outro-short], [outro-medium]")
+            else:
+                new_para_str = struct_tag
+        paragraphs_norm.append(new_para_str)
+    lyric_norm = " ; ".join(paragraphs_norm)
+
+    # format prompt 
     if prompt_audio is not None:
         genre = None
         description = None
@@ -84,13 +98,13 @@ def generate_song(lyric, description=None, prompt_audio=None, genre=None, cfg_co
     progress(0.0, "Start Generation")
     start = time.time()
     
-    audio_data = MODEL(lyric, description, prompt_audio, genre, op.join(APP_DIR, "ckpt/prompt.pt"), params).cpu().permute(1, 0).float().numpy()
+    audio_data = MODEL(lyric_norm, description, prompt_audio, genre, op.join(APP_DIR, "ckpt/prompt.pt"), params).cpu().permute(1, 0).float().numpy()
 
     end = time.time()
     
     # 创建输入配置的JSON
     input_config = {
-        "lyric": lyric,
+        "lyric": lyric_norm,
         "genre": genre,
         "prompt_audio": prompt_audio,
         "description": description,
@@ -125,8 +139,8 @@ lyrics
 '''
 1. One paragraph represents one segments, starting with a structure tag and ending with a blank line
 2. One line represents one sentence, punctuation is not recommended inside the sentence
-3. The following segments should not contain lyrics: [intro-short] , [intro-medium] , [inst-short] , [inst-medium] , [outro-short] , [outro-medium]
-4. The following segments require lyrics: [verse] , [chorus] , [bridge]
+3. The following segments should not contain lyrics: [intro-short], [intro-medium], [inst-short], [inst-medium], [outro-short], [outro-medium]
+4. The following segments require lyrics: [verse], [chorus], [bridge]
 """
             )
 
@@ -186,7 +200,7 @@ lyrics
         
         with gr.Column():
             output_audio = gr.Audio(label="Generated Song", type="numpy")
-            output_json = gr.JSON(label="Input Configuration")
+            output_json = gr.JSON(label="Generated Info")
     
         # # 示例按钮
         # examples = gr.Examples(