diff --git a/GPT_SoVITS/inference_webui.py b/GPT_SoVITS/inference_webui.py
index b588852f..fecbc57f 100644
--- a/GPT_SoVITS/inference_webui.py
+++ b/GPT_SoVITS/inference_webui.py
@@ -1245,6 +1245,34 @@ def _read_name2text_for_webui(logs_dir: Path) -> dict[str, dict[str, str]]:
     return output
 
 
+def _read_emotion_map_for_webui(model_name: str) -> dict[str, str]:
+    """读取 ASR 标注 .list 中的 emotion 列，返回 {audio_name|stem: emotion}。
+
+    扫描 output/asr_opt/<model_name>/ 下所有 *.list（4/5/6 列均兼容）。
+    无 emotion 列或文件不存在时返回空 dict。
+    """
+    from tools.list_metadata import parse_list_line
+
+    emotion_map: dict[str, str] = {}
+    list_dir = Path("output") / "asr_opt" / model_name
+    if not list_dir.exists():
+        return emotion_map
+    for list_path in sorted(list_dir.glob("*.list")):
+        try:
+            for line in list_path.read_text(encoding="utf-8", errors="ignore").splitlines():
+                item = parse_list_line(line)
+                if item is None or not item.emotion:
+                    continue
+                name = os.path.basename(item.wav_path.strip("\"'"))
+                if not name:
+                    continue
+                emotion_map[name] = item.emotion
+                emotion_map[Path(name).stem] = item.emotion
+        except Exception:
+            continue
+    return emotion_map
+
+
 def on_model_name_change(model_name: str):
     """模型名变化时，加载该模型的训练样本列表到下拉框与预览播放器。"""
     if not model_name:
@@ -1256,10 +1284,16 @@ def on_model_name_change(model_name: str):
     samples: list[tuple[str, str]] = []
     if wav_dir.exists():
         name2text = _read_name2text_for_webui(logs_dir)
+        emotion_map = _read_emotion_map_for_webui(model_name)
         for f in sorted(wav_dir.iterdir()):
             if f.is_file() and f.suffix.lower() in (".wav", ".mp3", ".flac"):
                 text = name2text.get(f.name, {}).get("text", "") or name2text.get(f.stem, {}).get("text", "")
-                label = f"{f.name}" + (f" | {text[:30]}" if text else "")
+                emotion = emotion_map.get(f.name, "") or emotion_map.get(f.stem, "")
+                label = f"{f.name}"
+                if text:
+                    label += f" | {text[:30]}"
+                if emotion:
+                    label += f" | 【{emotion}】"
                 samples.append((label, str(f)))
     choices = [s[0] for s in samples]
     value = samples[0][0] if samples else ""
@@ -1268,28 +1302,33 @@ def on_model_name_change(model_name: str):
 
 
 def on_ref_sample_change(sample_label: str, model_name: str):
-    """训练样本选择变化时，更新预览播放器。"""
+    """训练样本选择变化时，更新预览播放器与情绪/括注文本框。"""
     if not sample_label or not model_name:
-        return gr.Audio(value=None)
+        return gr.Audio(value=None), ""
     from config import exp_root
 
     wav_name = sample_label.split(" | ")[0]
     audio_path = Path(exp_root) / model_name / "5-wav32k" / wav_name
-    return gr.Audio(value=str(audio_path) if audio_path.exists() else None)
+    emotion_map = _read_emotion_map_for_webui(model_name)
+    emotion = emotion_map.get(wav_name, "") or emotion_map.get(Path(wav_name).stem, "")
+    return gr.Audio(value=str(audio_path) if audio_path.exists() else None), emotion
 
 
 def on_apply_sample(sample_label: str, model_name: str):
-    """应用选中样本：写入参考音频路径和参考文本。"""
+    """应用选中样本：写入参考音频路径、参考文本、情绪/括注。"""
     if not sample_label or not model_name:
-        return None, "", gr.Dropdown()
+        return None, "", gr.Dropdown(), ""
     from config import exp_root
 
     wav_name = sample_label.split(" | ")[0]
     audio_path = Path(exp_root) / model_name / "5-wav32k" / wav_name
     logs_dir = Path(exp_root) / model_name
     name2text = _read_name2text_for_webui(logs_dir)
+    emotion_map = _read_emotion_map_for_webui(model_name)
     text = name2text.get(wav_name, {}).get("text", "") or name2text.get(Path(wav_name).stem, {}).get("text", "")
-    return (str(audio_path) if audio_path.exists() else None), text, gr.Dropdown()
+    emotion = emotion_map.get(wav_name, "") or emotion_map.get(Path(wav_name).stem, "")
+    audio_out = str(audio_path) if audio_path.exists() else None
+    return audio_out, text, gr.Dropdown(), emotion
 
 
 def _scan_model_weights_for_webui() -> dict[str, dict[str, list[str]]]:
@@ -1412,6 +1451,12 @@ with gr.Blocks(title="GPT-SoVITS WebUI", analytics_enabled=False, js=js, css=css
                     )
                 )
                 prompt_text = gr.Textbox(label=i18n("参考音频的文本"), value="", lines=5, max_lines=5, scale=1)
+                ref_emotion_text = gr.Textbox(
+                    label=i18n("情绪/括注（来自训练样本，无数据留空）"),
+                    value="",
+                    interactive=False,
+                    scale=1,
+                )
             with gr.Column(scale=14):
                 prompt_language = gr.Dropdown(
                     label=i18n("参考音频的语种"),
@@ -1584,12 +1629,12 @@ with gr.Blocks(title="GPT-SoVITS WebUI", analytics_enabled=False, js=js, css=css
         ref_sample_dropdown.change(
             fn=on_ref_sample_change,
             inputs=[ref_sample_dropdown, model_name_dropdown],
-            outputs=[ref_sample_player],
+            outputs=[ref_sample_player, ref_emotion_text],
         )
         apply_sample_btn.click(
             fn=on_apply_sample,
             inputs=[ref_sample_dropdown, model_name_dropdown],
-            outputs=[inp_ref, prompt_text, ref_sample_dropdown],
+            outputs=[inp_ref, prompt_text, ref_sample_dropdown, ref_emotion_text],
         )
         auto_select_weights_btn.click(
             fn=on_auto_select_weights,
diff --git a/GPT_SoVITS/prepare_datasets/1-get-text.py b/GPT_SoVITS/prepare_datasets/1-get-text.py
index 8d83e79a..85db31c5 100644
--- a/GPT_SoVITS/prepare_datasets/1-get-text.py
+++ b/GPT_SoVITS/prepare_datasets/1-get-text.py
@@ -20,6 +20,7 @@ import os.path
 from text.cleaner import clean_text
 from transformers import AutoModelForMaskedLM, AutoTokenizer
 from tools.my_utils import clean_path
+from tools.list_metadata import parse_list_line
 
 # inp_text=sys.argv[1]
 # inp_wav_dir=sys.argv[2]
@@ -126,7 +127,12 @@ if os.path.exists(txt_path) == False:
     }
     for line in lines[int(i_part) :: int(all_parts)]:
         try:
-            wav_name, spk_name, language, text = line.split("|")
+            # 兼容 4/5/6 列 .list：parse_list_line 只取训练所需前 4 列，
+            # 额外的 emotion/remark 列（proplus-hc-dev 分支）被静默忽略，不影响训练。
+            item = parse_list_line(line)
+            if item is None:
+                raise ValueError("list 行列数不足，需至少 4 列：wav_path|speaker_name|language|text")
+            wav_name, language, text = item.wav_path, item.language, item.text
             # todo.append([name,text,"zh"])
             if language in language_v1_to_language_v2.keys():
                 todo.append([wav_name, text, language_v1_to_language_v2.get(language, language)])
diff --git a/api_v2.py b/api_v2.py
index b08c08b9..5ead07d4 100644
--- a/api_v2.py
+++ b/api_v2.py
@@ -637,6 +637,36 @@ def _read_name2text(logs_dir: Path) -> dict[str, dict[str, str]]:
     return output
 
 
+def _read_emotion_map(model_name: str) -> dict[str, str]:
+    """读取 ASR 标注 .list 中的 emotion 列，返回 {audio_name|stem: emotion}。
+
+    扫描 output/asr_opt/<model_name>/ 下所有 *.list（兼容 4/5/6 列）。
+    兼容 proplus-hc-dev 分支训练数据：emotion 为第 5 列，缺省为空串。
+    无 .list 或无 emotion 列时返回空 dict。
+    """
+    from tools.list_metadata import parse_list_line
+
+    emotion_map: dict[str, str] = {}
+    list_dir = Path("output") / "asr_opt" / model_name
+    if not list_dir.exists():
+        return emotion_map
+    for list_path in sorted(list_dir.glob("*.list")):
+        try:
+            for line in list_path.read_text(encoding="utf-8", errors="ignore").splitlines():
+                item = parse_list_line(line)
+                if item is None or not item.emotion:
+                    continue
+                name = os.path.basename(item.wav_path.strip("\"'"))
+                if not name:
+                    continue
+                emotion_map[name] = item.emotion
+                emotion_map[Path(name).stem] = item.emotion
+        except Exception:
+            continue
+    return emotion_map
+
+
+
 def _safe_logs_subpath(model_name: str, *parts: str) -> Path | None:
     """解析 logs/<model_name>/<parts...> 路径并做目录穿越校验。
 
@@ -706,6 +736,7 @@ async def list_model_samples(model_name: str):
     if logs_abs is None or not logs_abs.exists():
         return JSONResponse(status_code=404, content={"message": f"model '{model_name}' not found"})
     name2text = _read_name2text(logs_abs)
+    emotion_map = _read_emotion_map(model_name)
     wav_dir = logs_abs / "5-wav32k"
     samples: list[dict] = []
     if wav_dir.exists():
@@ -718,11 +749,13 @@ async def list_model_samples(model_name: str):
             if entry is None:
                 continue  # 与 name2text 取交集，只返回有标注的样本
             rel = os.path.relpath(f, now_dir)
+            emotion = emotion_map.get(f.name, "") or emotion_map.get(f.stem, "")
             samples.append({
                 "audio_name": f.name,
                 "audio_path": rel.replace(os.sep, "/"),
                 "text": entry["text"],
                 "lang": entry["lang"],
+                "emotion": emotion,
             })
     return {"model_name": model_name, "samples": samples, "total": len(samples)}
 
diff --git a/tools/list_metadata.py b/tools/list_metadata.py
new file mode 100644
index 00000000..eb232945
--- /dev/null
+++ b/tools/list_metadata.py
@@ -0,0 +1,61 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+
+@dataclass(frozen=True)
+class ListLine:
+    """一条 `.list` 标注行（兼容 4/5/6 列）。
+
+    列定义: wav_path|speaker_name|language|text[|emotion|remark]
+    训练只用前 4 列（training_fields）；emotion / remark 为元数据，
+    不参与模型训练，用于标注与参考音频展示。
+    """
+
+    wav_path: str
+    speaker_name: str
+    language: str
+    text: str
+    emotion: str = ""
+    remark: str = ""
+
+    @property
+    def training_fields(self):
+        return [self.wav_path, self.speaker_name, self.language, self.text]
+
+
+def _clean(value):
+    return str(value or "").strip()
+
+
+def parse_list_line(line):
+    """解析一行 `.list`。列数 < 4 返回 None；emotion/remark 缺省为空串。"""
+    parts = str(line or "").rstrip("\r\n").split("|")
+    if len(parts) < 4:
+        return None
+    return ListLine(
+        wav_path=_clean(parts[0]),
+        speaker_name=_clean(parts[1]),
+        language=_clean(parts[2]),
+        text=_clean(parts[3]),
+        emotion=_clean(parts[4]) if len(parts) >= 5 else "",
+        remark=_clean(parts[5]) if len(parts) >= 6 else "",
+    )
+
+
+def format_list_line(item):
+    """把 ListLine 格式化为 6 列 `.list` 行（emotion/remark 为空也保留分隔符）。
+
+    注意：写出 6 列以保证与 proplus-hc-dev 分支互操作，主分支训练侧
+    （1-get-text.py）已用 parse_list_line 只消费前 4 列，多余列不影响训练。
+    """
+    return "|".join(
+        [
+            str(item.wav_path or ""),
+            str(item.speaker_name or ""),
+            str(item.language or ""),
+            str(item.text or ""),
+            str(item.emotion or ""),
+            str(item.remark or ""),
+        ]
+    )
diff --git a/tools/subfix_webui.py b/tools/subfix_webui.py
index 51a7dfad..3ac0094e 100644
--- a/tools/subfix_webui.py
+++ b/tools/subfix_webui.py
@@ -32,6 +32,7 @@ g_batch = 10
 g_text_list = []
 g_audio_list = []
 g_checkbox_list = []
+g_emotion_list = []
 g_data_json = []
 
 
@@ -43,7 +44,13 @@ def reload_data(index, batch):
     datas = g_data_json[index : index + batch]
     output = []
     for d in datas:
-        output.append({g_json_key_text: d[g_json_key_text], g_json_key_path: d[g_json_key_path]})
+        output.append(
+            {
+                g_json_key_text: d[g_json_key_text],
+                g_json_key_path: d[g_json_key_path],
+                "emotion": d.get("emotion", ""),
+            }
+        )
     return output
 
 
@@ -52,26 +59,26 @@ def b_change_index(index, batch):
     g_index, g_batch = index, batch
     datas = reload_data(index, batch)
     output = []
+    # 文本框组
     for i, _ in enumerate(datas):
         output.append(
-            # gr.Textbox(
-            #     label=f"Text {i+index}",
-            #     value=_[g_json_key_text]#text
-            # )
             {"__type__": "update", "label": f"Text {i + index}", "value": _[g_json_key_text]}
         )
     for _ in range(g_batch - len(datas)):
+        output.append({"__type__": "update", "label": "Text", "value": ""})
+    # 情绪/括注框组（与文本框一一对应）
+    for i, _ in enumerate(datas):
         output.append(
-            # gr.Textbox(
-            #     label=f"Text",
-            #     value=""
-            # )
-            {"__type__": "update", "label": "Text", "value": ""}
+            {"__type__": "update", "label": f"情绪/括注 {i + index}", "value": _.get("emotion", "")}
         )
+    for _ in range(g_batch - len(datas)):
+        output.append({"__type__": "update", "label": "情绪/括注", "value": ""})
+    # 音频组
     for _ in datas:
         output.append(_[g_json_key_path])
     for _ in range(g_batch - len(datas)):
         output.append(None)
+    # 复选框组
     for _ in range(g_batch):
         output.append(False)
     return output
@@ -93,8 +100,12 @@ def b_previous_index(index, batch):
         return 0, *b_change_index(0, batch)
 
 
-def b_submit_change(*text_list):
+def b_submit_change(*args):
+    # 输入约定：[text×batch] + [emotion×batch]（由事件绑定顺序拼成）
     global g_data_json
+    half = g_batch
+    text_list = args[:half]
+    emotion_list = args[half:]
     change = False
     for i, new_text in enumerate(text_list):
         if g_index + i <= g_max_json_index:
@@ -102,6 +113,12 @@ def b_submit_change(*text_list):
             if g_data_json[g_index + i][g_json_key_text] != new_text:
                 g_data_json[g_index + i][g_json_key_text] = new_text
                 change = True
+    for i, new_emotion in enumerate(emotion_list):
+        if g_index + i <= g_max_json_index:
+            new_emotion = (new_emotion or "").strip()
+            if g_data_json[g_index + i].get("emotion", "") != new_emotion:
+                g_data_json[g_index + i]["emotion"] = new_emotion
+                change = True
     if change:
         b_save_file()
     return g_index, *b_change_index(g_index, g_batch)
@@ -226,13 +243,25 @@ def b_save_json():
 
 
 def b_save_list():
+    from tools.list_metadata import format_list_line, ListLine
+
     with open(g_load_file, "w", encoding="utf-8") as file:
         for data in g_data_json:
-            wav_path = data["wav_path"]
-            speaker_name = data["speaker_name"]
-            language = data["language"]
-            text = data["text"]
-            file.write(f"{wav_path}|{speaker_name}|{language}|{text}".strip() + "\n")
+            # 写出 6 列以保留 emotion/remark（proplus-hc-dev 互操作）；
+            # 主分支训练侧（1-get-text.py）只消费前 4 列，多余列不影响训练。
+            file.write(
+                format_list_line(
+                    ListLine(
+                        wav_path=data.get("wav_path", ""),
+                        speaker_name=data.get("speaker_name", ""),
+                        language=data.get("language", ""),
+                        text=data.get("text", ""),
+                        emotion=data.get("emotion", ""),
+                        remark=data.get("remark", ""),
+                    )
+                ).strip()
+                + "\n"
+            )
 
 
 def b_load_json():
@@ -245,17 +274,25 @@ def b_load_json():
 
 def b_load_list():
     global g_data_json, g_max_json_index
+    from tools.list_metadata import parse_list_line
+
     with open(g_load_file, "r", encoding="utf-8") as source:
         data_list = source.readlines()
         for _ in data_list:
-            data = _.split("|")
-            if len(data) == 4:
-                wav_path, speaker_name, language, text = data
-                g_data_json.append(
-                    {"wav_path": wav_path, "speaker_name": speaker_name, "language": language, "text": text.strip()}
-                )
-            else:
-                print("error line:", data)
+            item = parse_list_line(_)
+            if item is None:
+                print("error line:", _)
+                continue
+            g_data_json.append(
+                {
+                    "wav_path": item.wav_path,
+                    "speaker_name": item.speaker_name,
+                    "language": item.language,
+                    "text": item.text,
+                    "emotion": item.emotion,
+                    "remark": item.remark,
+                }
+            )
         g_max_json_index = len(g_data_json) - 1
 
 
@@ -338,9 +375,16 @@ if __name__ == "__main__":
                         text = gr.Textbox(label="Text", visible=True, scale=5)
                         audio_output = gr.Audio(label="Output Audio", visible=True, scale=5)
                         audio_check = gr.Checkbox(label="Yes", show_label=True, info="Choose Audio", scale=1)
+                        emotion = gr.Textbox(
+                            label="情绪/括注",
+                            visible=True,
+                            scale=3,
+                            placeholder="可选：台词对应的表演情绪或括注",
+                        )
                         g_text_list.append(text)
                         g_audio_list.append(audio_output)
                         g_checkbox_list.append(audio_check)
+                        g_emotion_list.append(emotion)
 
         with gr.Row():
             batchsize_slider = gr.Slider(
@@ -356,15 +400,16 @@ if __name__ == "__main__":
                 index_slider,
                 batchsize_slider,
             ],
-            outputs=[*g_text_list, *g_audio_list, *g_checkbox_list],
+            outputs=[*g_text_list, *g_emotion_list, *g_audio_list, *g_checkbox_list],
         )
 
         btn_submit_change.click(
             b_submit_change,
             inputs=[
                 *g_text_list,
+                *g_emotion_list,
             ],
-            outputs=[index_slider, *g_text_list, *g_audio_list, *g_checkbox_list],
+            outputs=[index_slider, *g_text_list, *g_emotion_list, *g_audio_list, *g_checkbox_list],
         )
 
         btn_previous_index.click(
@@ -373,7 +418,7 @@ if __name__ == "__main__":
                 index_slider,
                 batchsize_slider,
             ],
-            outputs=[index_slider, *g_text_list, *g_audio_list, *g_checkbox_list],
+            outputs=[index_slider, *g_text_list, *g_emotion_list, *g_audio_list, *g_checkbox_list],
         )
 
         btn_next_index.click(
@@ -382,25 +427,25 @@ if __name__ == "__main__":
                 index_slider,
                 batchsize_slider,
             ],
-            outputs=[index_slider, *g_text_list, *g_audio_list, *g_checkbox_list],
+            outputs=[index_slider, *g_text_list, *g_emotion_list, *g_audio_list, *g_checkbox_list],
         )
 
         btn_delete_audio.click(
             b_delete_audio,
             inputs=[*g_checkbox_list],
-            outputs=[index_slider, *g_text_list, *g_audio_list, *g_checkbox_list],
+            outputs=[index_slider, *g_text_list, *g_emotion_list, *g_audio_list, *g_checkbox_list],
         )
 
         btn_merge_audio.click(
             b_merge_audio,
             inputs=[interval_slider, *g_checkbox_list],
-            outputs=[index_slider, *g_text_list, *g_audio_list, *g_checkbox_list],
+            outputs=[index_slider, *g_text_list, *g_emotion_list, *g_audio_list, *g_checkbox_list],
         )
 
         btn_audio_split.click(
             b_audio_split,
             inputs=[splitpoint_slider, *g_checkbox_list],
-            outputs=[index_slider, *g_text_list, *g_audio_list, *g_checkbox_list],
+            outputs=[index_slider, *g_text_list, *g_emotion_list, *g_audio_list, *g_checkbox_list],
         )
 
         btn_invert_selection.click(b_invert_selection, inputs=[*g_checkbox_list], outputs=[*g_checkbox_list])
@@ -413,7 +458,7 @@ if __name__ == "__main__":
                 index_slider,
                 batchsize_slider,
             ],
-            outputs=[*g_text_list, *g_audio_list, *g_checkbox_list],
+            outputs=[*g_text_list, *g_emotion_list, *g_audio_list, *g_checkbox_list],
         )
 
     demo.launch(