Update app.py

app.py

@@ -41,11 +41,11 @@ async def tts_interface(text, voice, rate_percentage, pitch_hz):
 def play_saved_audio(audio_file):
     return audio_file
 
-async def podcast_produce(script_list, voice, rate_percentage, pitch_hz, bgm_file, podcast_title, podcast_desc):
+async def podcast_produce(scripts, voice, rate_percentage, pitch_hz, bgm_file, podcast_title, podcast_desc):
     rate = f"{'+' if rate_percentage >= 0 else ''}{rate_percentage}%"
     pitch = f"{'+' if pitch_hz >= 0 else ''}{pitch_hz}Hz"
     audio_segments = []
-    for idx, text in enumerate(script_list):
+    for idx, text in enumerate(scripts):
         if text.strip():
             temp_audio = generate_unique_filename(PODCAST_DIR, prefix=f"segment{idx}")
             communicate = edge_tts.Communicate(text, voice, rate=rate, pitch=pitch)
@@ -55,7 +55,7 @@ async def podcast_produce(script_list, voice, rate_percentage, pitch_hz, bgm_fil
     if not audio_segments:
         return None
     podcast_audio = sum(audio_segments)
-    if bgm_file is not None and os.path.isfile(bgm_file.name):
+    if bgm_file is not None and hasattr(bgm_file, "name") and os.path.isfile(bgm_file.name):
         bgm = AudioSegment.from_file(bgm_file.name).apply_gain(-10)
         bgm = bgm[:len(podcast_audio)]
         podcast_audio = podcast_audio.overlay(bgm)
@@ -66,24 +66,30 @@ async def podcast_produce(script_list, voice, rate_percentage, pitch_hz, bgm_fil
         f.write(f"Title: {podcast_title}\nDescription: {podcast_desc}\n")
     return podcast_file
 
+def clear_textbox():
+    return ""
+
+def clear_paragraphs():
+    return [""]
+
 def add_paragraph(paragraphs):
+    paragraphs = paragraphs.copy()
     paragraphs.append("")
     return paragraphs
 
 def remove_paragraph(paragraphs):
+    paragraphs = paragraphs.copy()
     if len(paragraphs) > 1:
         paragraphs.pop()
     return paragraphs
 
-def clear_paragraphs():
-    return [""]
-
-def clear_textbox():
-    return ""
+def update_paragraphs_ui(paragraphs):
+    # 回傳一組 Textbox 元件
+    return [gr.Textbox(value=p, label=f"段落{i+1}內容", lines=3, interactive=True) for i, p in enumerate(paragraphs)]
 
-def render_paragraphs(paragraphs):
-    # 回傳一組 Textbox 的 dict，方便動態更新
-    return {f"para_{i}": p for i, p in enumerate(paragraphs)}
+def collect_paragraphs(*args):
+    # 收集所有段落內容
+    return list(args)
 
 async def main():
     voices = await get_voices()
@@ -113,57 +119,35 @@ async def main():
 
         with gr.Tab("播客製作"):
             gr.Markdown("### 📝 多段腳本輸入（可自由增減段落）")
-
             paragraphs_state = gr.State([""])
+            paragraphs_container = gr.Column()
 
-            # 動態產生多個 Textbox，並用 dictionary 存放元件
-            paragraph_textboxes = {}
+            # 初始渲染
+            paragraph_boxes = update_paragraphs_ui([""])
+            for tb in paragraph_boxes:
+                paragraphs_container.append(tb)
 
-            def render_paragraph_boxes(paragraphs):
-                boxes = {}
-                for i, p in enumerate(paragraphs):
-                    boxes[f"para_{i}"] = gr.Textbox(value=p, label=f"段落{i+1}內容", lines=3)
-                return boxes
-
-            paragraph_textboxes = render_paragraph_boxes(paragraphs_state.value)
-
-            # 按鈕
             add_btn = gr.Button("新增段落")
             remove_btn = gr.Button("刪除段落")
             clear_all_btn = gr.Button("全部清空")
 
-            # 更新段落文字
-            def update_paragraphs(*texts):
-                return list(texts)
-
-            # 新增段落
             def on_add(paragraphs):
-                paragraphs.append("")
-                return paragraphs
+                new_paragraphs = add_paragraph(paragraphs)
+                return new_paragraphs, gr.update(components=update_paragraphs_ui(new_paragraphs))
 
-            # 刪除段落
             def on_remove(paragraphs):
-                if len(paragraphs) > 1:
-                    paragraphs.pop()
-                return paragraphs
+                new_paragraphs = remove_paragraph(paragraphs)
+                return new_paragraphs, gr.update(components=update_paragraphs_ui(new_paragraphs))
 
-            # 全部清空
             def on_clear():
-                return [""]
+                new_paragraphs = clear_paragraphs()
+                return new_paragraphs, gr.update(components=update_paragraphs_ui(new_paragraphs))
 
-            add_btn.click(on_add, inputs=paragraphs_state, outputs=paragraphs_state)
-            remove_btn.click(on_remove, inputs=paragraphs_state, outputs=paragraphs_state)
-            clear_all_btn.click(on_clear, outputs=paragraphs_state)
-
-            # 監聽段落文字改變，更新狀態
-            def on_text_change(*texts):
-                return list(texts)
-
-            # 將段落文字元件放入列表，方便傳入合成函數
-            def get_paragraph_values(*args):
-                return list(args)
+            add_btn.click(on_add, inputs=paragraphs_state, outputs=[paragraphs_state, paragraphs_container])
+            remove_btn.click(on_remove, inputs=paragraphs_state, outputs=[paragraphs_state, paragraphs_container])
+            clear_all_btn.click(on_clear, outputs=[paragraphs_state, paragraphs_container])
 
-            # 播客參數輸入
+            # 參數設定
             voice_input2 = gr.Dropdown(voices, label="選擇語音", value="zh-CN-XiaoxiaoNeural")
             rate_input2 = gr.Slider(-50, 50, value=0, step=1, label="語速調整 (%)")
             pitch_input2 = gr.Slider(-50, 50, value=0, step=1, label="音高調整 (Hz)")
@@ -173,28 +157,12 @@ async def main():
             podcast_btn = gr.Button("生成播客")
             podcast_output = gr.Audio(type="filepath", label="生成的播客音檔")
 
-            # 動態建立段落輸入元件區塊
-            with gr.Column() as paragraphs_container:
-                def build_paragraph_boxes(paragraphs):
-                    boxes = []
-                    for i, p in enumerate(paragraphs):
-                        tb = gr.Textbox(value=p, label=f"段落{i+1}內容", lines=3)
-                        boxes.append(tb)
-                    return boxes
-
-                paragraph_boxes = build_paragraph_boxes(paragraphs_state.value)
-
-            # 重新渲染段落輸入框
-            def rerender_paragraph_boxes(paragraphs):
-                nonlocal paragraph_boxes
-                paragraph_boxes = build_paragraph_boxes(paragraphs)
-                return paragraph_boxes
-
-            paragraphs_state.change(rerender_paragraph_boxes, inputs=paragraphs_state, outputs=paragraph_boxes)
+            def gather_scripts(*args):
+                return list(args)
 
-            # 播客合成按鈕事件
             def on_podcast_btn_click(*args):
-                n = len(paragraph_boxes)
+                # args: 段落內容 + 參數
+                n = len(paragraphs_state.value)
                 scripts = list(args[:n])
                 voice = args[n]
                 rate = args[n+1]
@@ -204,8 +172,15 @@ async def main():
                 desc = args[n+5]
                 return asyncio.run(podcast_produce(scripts, voice, rate, pitch, bgm, title, desc))
 
-            inputs = paragraph_boxes + [voice_input2, rate_input2, pitch_input2, bgm_input, podcast_title, podcast_desc]
-            podcast_btn.click(on_podcast_btn_click, inputs=inputs, outputs=podcast_output)
+            # 這裡需要用最新的段落數來組合 inputs
+            def get_inputs():
+                return [tb for tb in paragraphs_container.children] + [voice_input2, rate_input2, pitch_input2, bgm_input, podcast_title, podcast_desc]
+
+            podcast_btn.click(
+                fn=on_podcast_btn_click,
+                inputs=lambda: get_inputs(),
+                outputs=podcast_output
+            )
 
         with gr.Tab("檢視已儲存播客"):
             podcast_files = gr.Dropdown(list_saved_podcasts(), label="選擇已儲存播客檔案", interactive=True)