Update app.py

app.py

@@ -83,18 +83,25 @@ def split_audio_if_needed(path):
     return files
 
 def transcribe_core(path, model):
-    # ✅ 處理 iPhone LINE 語音（mp4 audio-only，不轉檔，只改副檔名）
-    if path.lower().endswith(".mp4"):
+    # ✅ iPhone LINE 語音（mp4 audio-only）— 不轉檔，只複製改副檔名
+    if path and path.lower().endswith(".mp4"):
         fixed_path = path[:-4] + ".m4a"
-        shutil.copy(path, fixed_path)  # 複製一份改名，不耗資源
-        path = fixed_path
-        print("🔧 已自動修正 mp4 → m4a")
+        try:
+            shutil.copy(path, fixed_path)
+            path = fixed_path
+            print("🔧 已自動修正 mp4 → m4a")
+        except Exception as e:
+            print(f"⚠️ mp4→m4a 複製失敗：{e}，改用原檔嘗試")
 
     chunks = split_audio_if_needed(path)
     txts = []
     for f in chunks:
         with open(f, "rb") as af:
-            res = client.audio.transcriptions.create(model=model, file=af, response_format="text")
+            res = client.audio.transcriptions.create(
+                model=model,
+                file=af,
+                response_format="text"
+            )
         txts.append(res)
     full = "\n".join(txts)
     res = client.chat.completions.create(
@@ -108,18 +115,38 @@ def transcribe_core(path, model):
 # ========================
 # 💬 主流程
 # ========================
-def transcribe_with_password(session_id, password, file, model_choice):
+def _normalize_upload_path(file_input):
+    """兼容 gr.File 回傳型態：字串 / 物件 / dict / list"""
+    if not file_input:
+        return None
+    if isinstance(file_input, str):
+        return file_input
+    if isinstance(file_input, list) and file_input:
+        return _normalize_upload_path(file_input[0])
+    # gradio 有時傳 UploadedFile 物件或 dict
+    path = getattr(file_input, "name", None)
+    if not path and isinstance(file_input, dict):
+        path = file_input.get("name") or file_input.get("path")
+    return path
+
+def transcribe_with_password(session_id, password, file_input, model_choice):
+    # 修正注音輸入造成的隱藏字元
     password = password.strip().replace(" ", "").replace("\u200b", "")
+
     locked_flag, msg = check_lock(session_id)
     if locked_flag:
         return msg, "", ""
+
     if password != PASSWORD:
         cnt, msg2 = record_failed_attempt(session_id)
         return msg2 or f"密碼錯誤（第 {cnt} 次）", "", ""
-    if not file:
-        return "請上傳音訊檔。", "", ""
+
+    path = _normalize_upload_path(file_input)
+    if not path or not os.path.exists(path):
+        return "找不到上傳檔案，請重新選擇。", "", ""
+
     clear_attempts(session_id)
-    full, summ = transcribe_core(file, model_choice)
+    full, summ = transcribe_core(path, model_choice)
     return "✅ 轉錄完成", full, summ
 
 def ask_about_transcript(full_text, q):
@@ -129,14 +156,17 @@ def ask_about_transcript(full_text, q):
         return "請輸入問題"
     prompt = f"以下是轉錄內容：\n{full_text}\n\n問題：{q}\n請用繁體中文回答。"
     res = client.chat.completions.create(
-        model="gpt-4o-mini", messages=[{"role":"user","content":prompt}], temperature=0.6)
+        model="gpt-4o-mini",
+        messages=[{"role":"user","content":prompt}],
+        temperature=0.6,
+    )
     return res.choices[0].message.content.strip()
 
 # ========================
 # 🌐 Gradio 介面
 # ========================
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.Markdown("## 🎧 語音轉錄與摘要工具（私人API勿轉傳）")
+    gr.Markdown("## 🎧 語音轉錄與摘要工具（私人API勿轉傳｜支援 iPhone LINE .mp4）")
 
     session_state = gr.State(value=None)
 
@@ -153,7 +183,13 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
             label="選擇模型"
         )
 
-    audio_input = gr.Audio(type="filepath", label="上傳音訊 (.m4a, .aac, .wav, .mp4)")
+    # ✅ 用 File 而不是 Audio，允許 video/mp4 以及常見音訊附檔名
+    file_input = gr.File(
+        label="上傳音訊 / LINE 語音檔（支援 .m4a, .aac, .wav, .mp4）",
+        file_count="single",
+        file_types=["audio", ".mp4", ".m4a", ".aac", ".wav"]
+    )
+
     transcribe_btn = gr.Button("開始轉錄與摘要 🚀")
     status_box = gr.Textbox(label="狀態", interactive=False)
     transcript_box = gr.Textbox(label="完整轉錄文字", lines=10)
@@ -174,12 +210,12 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
 
     transcribe_btn.click(
         transcribe_with_password,
-        [session_state, password_input, audio_input, model_choice],
+        [session_state, password_input, file_input, model_choice],
         [status_box, transcript_box, summary_box],
     )
     ask_btn.click(ask_about_transcript, [transcript_box, user_q], [ai_reply])
 
-    # ✅ 正確的 JS 複製寫法 (Gradio 5.x)
+    # ✅ 複製（Gradio 5.x）
     copy_js = """
     async (text) => {
         try {