diff --git a/.gitignore b/.gitignore
index 0db34f7..1dd404e 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,12 +1,31 @@
-venv/
-__pycache__
-*.egg-info
-*.DS_Store
+# Development Tools.
+.mypy_cache/
+.ruff_cache/
+__pycache__/
 .idea/
 .vscode/
-checkpoints/*
-outputs/
-build/
+
+# Environments.
+.venv*/
+venv*/
+conda_env*/
+
+# Python Bytecode.
 *.py[cod]
+
+# Distribution/Packaging.
+/build/
+/dist/
 *.egg-info/
-.venv
+.pypirc
+
+# Operating System Junk.
+*.DS_Store
+Thumbs.db
+desktop.ini
+
+# IndexTTS.
+/cache/
+/checkpoints/*
+!/checkpoints/*.yaml
+/outputs/
diff --git a/README.md b/README.md
index 3465bc4..c4b90e2 100644
--- a/README.md
+++ b/README.md
@@ -95,25 +95,31 @@ The key contributions of **indextts2** are summarized as follows:
 
 ## Usage Instructions
 ### Environment Setup
+
 1. Download this repository:
 ```bash
 git clone https://github.com/index-tts/index-tts.git
 git lfs pull # fetch example files
 ```
-2. Install dependencies:
-We use `uv` to manage the project’s dependency environment.
+
+2. Install the [uv](https://docs.astral.sh/uv/getting-started/installation/) package
+   manager. It is *required* for a reliable, modern installation environment.
+
+3. Install dependencies:
+
+We use `uv` to manage the project's dependency environment.
 ```bash
 uv sync
 ```
 
-3. Download models:
+4. Download models:
 
-Download by `huggingface-cli`:
+Download via `huggingface-cli`:
 ```bash
 hf download IndexTeam/IndexTTS-2 --local-dir=checkpoints
 ```
 
-Or download by `modelscope`
+Or download via `modelscope`
 ```bash
 modelscope download --model IndexTeam/IndexTTS-2 --local_dir checkpoints
 ```
@@ -158,7 +164,7 @@ tts.infer(spk_audio_prompt='examples/voice_07.wav', text=text, output_path="gen.
 ```
 
 
-4. It’s also possible to omit the emotional reference audio and instead provide an 8-float list specifying the intensity of each base emotion (Happy | Angery | Sad | Fear | Hate | Low | Surprise | Neutral). You can additionally control the `use_random` parameter to decide whether to introduce stochasticity during inference; the default is `False`, and setting it to `True` increases randomness:
+4. It's also possible to omit the emotional reference audio and instead provide an 8-float list specifying the intensity of each base emotion (Happy | Angery | Sad | Fear | Hate | Low | Surprise | Neutral). You can additionally control the `use_random` parameter to decide whether to introduce stochasticity during inference; the default is `False`, and setting it to `True` increases randomness:
 ```python
 from indextts.infer_v2 import IndexTTS2
 tts = IndexTTS2(cfg_path="checkpoints/config.yaml", model_dir="checkpoints", is_fp16=False, use_cuda_kernel=False)
diff --git a/examples/cases.jsonl b/examples/cases.jsonl
index d688b8d..1b71672 100644
--- a/examples/cases.jsonl
+++ b/examples/cases.jsonl
@@ -1,4 +1,4 @@
-{"prompt_audio":"voice_01.wav","text":"Translate for me，what is a surprise！","emo_mode":0}
+{"prompt_audio":"voice_01.wav","text":"Translate for me, what is a surprise!","emo_mode":0}
 {"prompt_audio":"voice_02.wav","text":"The palace is strict, no false rumors, Lady Qi!","emo_mode":0}
 {"prompt_audio":"voice_03.wav","text":"这个呀，就是我们精心制作准备的纪念品，大家可以看到这个色泽和这个材质啊，哎呀多么的光彩照人。","emo_mode":0}
 {"prompt_audio":"voice_04.wav","text":"你就需要我这种专业人士的帮助，就像手无缚鸡之力的人进入雪山狩猎，一定需要最老练的猎人指导。","emo_mode":0}
diff --git a/indextts/infer_v2.py b/indextts/infer_v2.py
index 3ab0861..2f0f1d3 100644
--- a/indextts/infer_v2.py
+++ b/indextts/infer_v2.py
@@ -2,6 +2,8 @@ import os
 from subprocess import CalledProcessError
 
 os.environ['HF_HUB_CACHE'] = './checkpoints/hf_cache'
+import json
+import re
 import time
 import librosa
 import torch
@@ -80,23 +82,15 @@ class IndexTTS2:
         else:
             self.gpt.eval()
         print(">> GPT weights restored from:", self.gpt_path)
-        if self.is_fp16:
-            use_deepspeed = True
-            try:
-                import deepspeed
-            except (ImportError, OSError, CalledProcessError) as e:
-                use_deepspeed = False
-                print(f">> DeepSpeed加载失败，回退到标准推理: {e}")
 
-            self.gpt.post_init_gpt2_config(use_deepspeed=use_deepspeed, kv_cache=True, half=True)
-        else:
-            use_deepspeed = True
-            try:
-                import deepspeed
-            except (ImportError, OSError, CalledProcessError) as e:
-                use_deepspeed = False
-                print(f">> DeepSpeed加载失败，回退到标准推理: {e}")
-            self.gpt.post_init_gpt2_config(use_deepspeed=use_deepspeed, kv_cache=True, half=False)
+        use_deepspeed = True
+        try:
+            import deepspeed
+        except (ImportError, OSError, CalledProcessError) as e:
+            use_deepspeed = False
+            print(f">> DeepSpeed加载失败，回退到标准推理: {e}")
+
+        self.gpt.post_init_gpt2_config(use_deepspeed=use_deepspeed, kv_cache=True, half=self.is_fp16)
 
         if self.use_cuda_kernel:
             # preload the CUDA kernel for BigVGAN
@@ -315,8 +309,9 @@ class IndexTTS2:
             # assert emo_alpha == 1.0
             if emo_text is None:
                 emo_text = text
-            emo_dict, content = self.qwen_emo.inference(emo_text)
+            emo_dict = self.qwen_emo.inference(emo_text)
             print(emo_dict)
+            # convert ordered dict to list of vectors; the order is VERY important!
             emo_vector = list(emo_dict.values())
 
         if emo_vector is not None:
@@ -602,59 +597,52 @@ class QwenEmotion:
             device_map="auto"
         )
         self.prompt = "文本情感分类"
-        self.convert_dict = {
-            "愤怒": "angry",
+        self.cn_key_to_en = {
             "高兴": "happy",
-            "恐惧": "fear",
-            "反感": "hate",
+            "愤怒": "angry",
             "悲伤": "sad",
-            "低落": "low",
-            "惊讶": "surprise",
-            "自然": "neutral",
+            "恐惧": "afraid",
+            "反感": "disgusted",
+            # TODO: the "低落" (melancholic) emotion will always be mapped to
+            # "悲伤" (sad) by QwenEmotion's text analysis. it doesn't know the
+            # difference between those emotions even if user writes exact words.
+            # SEE: `self.melancholic_words` for current workaround.
+            "低落": "melancholic",
+            "惊讶": "surprised",
+            "自然": "calm",
+        }
+        self.desired_vector_order = ["高兴", "愤怒", "悲伤", "恐惧", "反感", "低落", "惊讶", "自然"]
+        self.melancholic_words = {
+            # emotion text phrases that will force QwenEmotion's "悲伤" (sad) detection
+            # to become "低落" (melancholic) instead, to fix limitations mentioned above.
+            "低落",
+            "melancholy",
+            "melancholic",
+            "depression",
+            "depressed",
+            "gloomy",
         }
-        self.backup_dict = {"happy": 0, "angry": 0, "sad": 0, "fear": 0, "hate": 0, "low": 0, "surprise": 0,
-                            "neutral": 1.0}
         self.max_score = 1.2
         self.min_score = 0.0
 
+    def clamp_score(self, value):
+        return max(self.min_score, min(self.max_score, value))
+
     def convert(self, content):
-        content = content.replace("\n", " ")
-        content = content.replace(" ", "")
-        content = content.replace("{", "")
-        content = content.replace("}", "")
-        content = content.replace('"', "")
-        parts = content.strip().split(',')
-        print(parts)
-        parts_dict = {}
-        desired_order = ["高兴", "愤怒", "悲伤", "恐惧", "反感", "低落", "惊讶", "自然"]
-        for part in parts:
-            key_value = part.strip().split(':')
-            if len(key_value) == 2:
-                parts_dict[key_value[0].strip()] = part
-        # 按照期望顺序重新排列
-        ordered_parts = [parts_dict[key] for key in desired_order if key in parts_dict]
-        parts = ordered_parts
-        if len(parts) != len(self.convert_dict):
-            return self.backup_dict
+        # generate emotion vector dictionary:
+        # - insert values in desired order (Python 3.7+ `dict` remembers insertion order)
+        # - convert Chinese keys to English
+        # - clamp all values to the allowed min/max range
+        # - use 0.0 for any values that were missing in `content`
+        emotion_dict = {
+            self.cn_key_to_en[cn_key]: self.clamp_score(content.get(cn_key, 0.0))
+            for cn_key in self.desired_vector_order
+        }
 
-        emotion_dict = {}
-        for part in parts:
-            key_value = part.strip().split(':')
-            if len(key_value) == 2:
-                try:
-                    key = self.convert_dict[key_value[0].strip()]
-                    value = float(key_value[1].strip())
-                    value = max(self.min_score, min(self.max_score, value))
-                    emotion_dict[key] = value
-                except Exception:
-                    continue
-
-        for key in self.backup_dict:
-            if key not in emotion_dict:
-                emotion_dict[key] = 0.0
-
-        if sum(emotion_dict.values()) <= 0:
-            return self.backup_dict
+        # default to a calm/neutral voice if all emotion vectors were empty
+        if all(val <= 0.0 for val in emotion_dict.values()):
+            print(">> no emotions detected; using default calm/neutral voice")
+            emotion_dict["calm"] = 1.0
 
         return emotion_dict
 
@@ -687,9 +675,30 @@ class QwenEmotion:
         except ValueError:
             index = 0
 
-        content = self.tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")
-        emotion_dict = self.convert(content)
-        return emotion_dict, content
+        content = self.tokenizer.decode(output_ids[index:], skip_special_tokens=True)
+
+        # decode the JSON emotion detections as a dictionary
+        try:
+            content = json.loads(content)
+        except json.decoder.JSONDecodeError:
+            # invalid JSON; fallback to manual string parsing
+            # print(">> parsing QwenEmotion response", content)
+            content = {
+                m.group(1): float(m.group(2))
+                for m in re.finditer(r'([^\s":.,]+?)"?\s*:\s*([\d.]+)', content)
+            }
+            # print(">> dict result", content)
+
+        # workaround for QwenEmotion's inability to distinguish "悲伤" (sad) vs "低落" (melancholic).
+        # if we detect any of the IndexTTS "melancholic" words, we swap those vectors
+        # to encode the "sad" emotion as "melancholic" (instead of sadness).
+        text_input_lower = text_input.lower()
+        if any(word in text_input_lower for word in self.melancholic_words):
+            # print(">> before vec swap", content)
+            content["悲伤"], content["低落"] = content.get("低落", 0.0), content.get("悲伤", 0.0)
+            # print(">>  after vec swap", content)
+
+        return self.convert(content)
 
 
 if __name__ == "__main__":
diff --git a/indextts/utils/front.py b/indextts/utils/front.py
index eefd19f..bc7262c 100644
--- a/indextts/utils/front.py
+++ b/indextts/utils/front.py
@@ -91,7 +91,7 @@ class TextNormalizer:
         import platform
         if self.zh_normalizer is not None and self.en_normalizer is not None:
             return
-        if platform.system() == "Darwin" or platform.system() == "Windows":
+        if platform.system() != "Linux":  # Mac and Windows
             from wetext import Normalizer
 
             self.zh_normalizer = Normalizer(remove_erhua=False, lang="zh", operator="tn")
diff --git a/pyproject.toml b/pyproject.toml
index 1e972cf..10bcd41 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,10 +1,13 @@
 [project]
 name = "index-tts"
-version = "0.1.0"
-description = "Add your description here"
+version = "2.0.0"
+description = "IndexTTS2: A Breakthrough in Emotionally Expressive and Duration-Controlled Auto-Regressive Zero-Shot Text-to-Speech"
+license = "Apache-2.0"
+license-files = ["LICEN[CS]E*", "INDEX_MODEL_LICENSE*"]
 readme = "README.md"
 requires-python = ">=3.10"
 dependencies = [
+    # IMPORTANT: Always run `uv lock` to resolve and update the lockfile after edits:
     "accelerate==1.8.1",
     "cn2an==0.5.22",
     "cython==3.0.7",
@@ -24,7 +27,7 @@ dependencies = [
     "numpy==1.26.2",
     "omegaconf>=2.3.0",
     "opencv-python==4.9.0.80",
-    "pandas==2.1.3",
+    "pandas==2.3.2",
     "safetensors==0.5.2",
     "sentencepiece>=0.2.1",
     "tensorboard==2.9.1",
@@ -32,10 +35,16 @@ dependencies = [
     "tokenizers==0.21.0",
     "tqdm>=4.67.1",
     "transformers==4.52.1",
-    "wetext>=0.0.9;sys_platform != 'linux'",
-    "WeTextProcessing;sys_platform == 'linux'",
+
+    # Use "wetext" on Windows/Mac, otherwise "WeTextProcessing" on Linux."
+    "wetext>=0.0.9; sys_platform != 'linux'",
+    "WeTextProcessing; sys_platform == 'linux'",
 ]
 
+[project.urls]
+Homepage = "https://github.com/index-tts/index-tts"
+Repository = "https://github.com/index-tts/index-tts.git"
+
 [tool.uv]
 extra-index-url = ["https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"]
 no-build-isolation-package = ["deepspeed"]
diff --git a/requirements.txt b/requirements.txt
deleted file mode 100644
index e17b10b..0000000
--- a/requirements.txt
+++ /dev/null
@@ -1,30 +0,0 @@
-accelerate==1.8.1
-descript-audiotools==0.7.2
-transformers==4.52.1
-tokenizers==0.21.0
-cn2an==0.5.22
-ffmpeg-python==0.2.0
-Cython==3.0.7
-g2p-en==2.1.0
-jieba==0.42.1
-json5==0.10.0
-keras==2.9.0
-numba==0.58.1
-numpy==1.26.2
-pandas==2.1.3
-matplotlib==3.8.2
-munch==4.0.0
-opencv-python==4.9.0.80
-tensorboard==2.9.1
-librosa==0.10.2.post1
-safetensors==0.5.2
-deepspeed==0.17.1
-modelscope==1.27.0
-omegaconf
-sentencepiece
-gradio
-tqdm
-textstat
-
-WeTextProcessing; platform_machine == "Linux"
-wetext; platform_system != "Linux"
diff --git a/tools/i18n/locale/en_US.json b/tools/i18n/locale/en_US.json
index 8ef3daf..620bb36 100644
--- a/tools/i18n/locale/en_US.json
+++ b/tools/i18n/locale/en_US.json
@@ -11,11 +11,11 @@
   "喜": "Happy",
   "怒": "Angry",
   "哀": "Sad",
-  "惧": "Fear",
-  "厌恶": "Hate",
-  "低落": "Low",
-  "惊喜": "Surprise",
-  "平静": "Neutral",
+  "惧": "Afraid",
+  "厌恶": "Disgusted",
+  "低落": "Melancholic",
+  "惊喜": "Surprised",
+  "平静": "Calm",
   "情感描述文本": "Emotion description",
   "请输入情感描述文本": "Please input emotion description",
   "高级生成参数设置": "Advanced generation parameter settings",
@@ -28,8 +28,8 @@
   "功能设置": "Settings",
   "分句设置": "Sentence segmentation settings",
   "参数会影响音频质量和生成速度": "Parameters below affect audio quality and generation speed",
-  "分句最大Token数": "Max tokens per sentence",
-  "建议80~200之间，值越大，分句越长；值越小，分句越碎；过小过大都可能导致音频质量不高": "Recommended between 80 and 200. The larger the value, the longer the sentences; the smaller the value, the more fragmented the sentences. Values that are too small or too large may lead to poor audio quality.",
+  "分句最大Token数": "Max tokens per generation segment",
+  "建议80~200之间，值越大，分句越长；值越小，分句越碎；过小过大都可能导致音频质量不高": "Recommended range: 80 - 200. Larger values require more VRAM but improves the flow of the speech, while lower values require less VRAM but means more fragmented sentences. Values that are too small or too large may lead to less coherent speech.",
   "预览分句结果": "Preview sentence segmentation result",
   "序号": "Index",
   "分句内容": "Content",
@@ -37,6 +37,8 @@
   "情感控制方式": "Emotion control method",
   "GPT2 采样设置": "GPT-2 Sampling Configuration",
   "参数会影响音频多样性和生成速度详见": "Influence both the diversity of the generated audio and the generation speed. For further details, refer to",
+  "是否进行采样": "Enable GPT-2 sampling",
+  "生成Token最大数量，过小导致音频被截断": "Maximum number of tokens to generate. If text exceeds this, the audio will be cut off.",
   "请上传情感参考音频": "Please upload emotion reference audio",
   "当前模型版本": "Current model version ",
   "请输入目标文本": "Please input text to synthesize",
diff --git a/tools/i18n/locale/zh_CN.json b/tools/i18n/locale/zh_CN.json
index 26a7596..95aaf56 100644
--- a/tools/i18n/locale/zh_CN.json
+++ b/tools/i18n/locale/zh_CN.json
@@ -36,5 +36,7 @@
   "Token数": "Token数",
   "情感控制方式": "情感控制方式",
   "GPT2 采样设置": "GPT2 采样设置",
-  "参数会影响音频多样性和生成速度详见": "参数会影响音频多样性和生成速度详见"
+  "参数会影响音频多样性和生成速度详见": "参数会影响音频多样性和生成速度详见",
+  "是否进行采样": "是否进行采样",
+  "生成Token最大数量，过小导致音频被截断": "生成Token最大数量，过小导致音频被截断"
 }
\ No newline at end of file
diff --git a/uv.lock b/uv.lock
index f9e9d09..2a2f555 100644
--- a/uv.lock
+++ b/uv.lock
@@ -1176,7 +1176,7 @@ requires-dist = [
     { name = "numpy", specifier = "==1.26.2" },
     { name = "omegaconf", specifier = ">=2.3.0" },
     { name = "opencv-python", specifier = "==4.9.0.80" },
-    { name = "pandas", specifier = "==2.1.3" },
+    { name = "pandas", specifier = "==2.3.2" },
     { name = "safetensors", specifier = "==0.5.2" },
     { name = "sentencepiece", specifier = ">=0.2.1" },
     { name = "tensorboard", specifier = "==2.9.1" },
@@ -2153,7 +2153,7 @@ wheels = [
 
 [[package]]
 name = "pandas"
-version = "2.1.3"
+version = "2.3.2"
 source = { registry = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple" }
 dependencies = [
     { name = "numpy" },
@@ -2161,26 +2161,42 @@ dependencies = [
     { name = "pytz" },
     { name = "tzdata" },
 ]
-sdist = { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/86/ff/662dde2193fc93b8547b073db20472b9676f944d907247a46c9c5bc45bfc/pandas-2.1.3.tar.gz", hash = "sha256:22929f84bca106921917eb73c1521317ddd0a4c71b395bcf767a106e3494209f", size = 4272855, upload-time = "2023-11-10T19:19:47.654Z" }
+sdist = { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/79/8e/0e90233ac205ad182bd6b422532695d2b9414944a280488105d598c70023/pandas-2.3.2.tar.gz", hash = "sha256:ab7b58f8f82706890924ccdfb5f48002b83d2b5a3845976a9fb705d36c34dcdb", size = 4488684, upload-time = "2025-08-21T10:28:29.257Z" }
 wheels = [
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/6a/81/711b32480f508dcf29cb501f61c4f1c3c72409e9168c6625145440c1c320/pandas-2.1.3-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:acf08a73b5022b479c1be155d4988b72f3020f308f7a87c527702c5f8966d34f", size = 11722181, upload-time = "2023-11-10T19:13:33.959Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/fc/85/ec986ea64f55013d8c669da657f0da86383a15668f9814be2001e08a4807/pandas-2.1.3-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:3cc4469ff0cf9aa3a005870cb49ab8969942b7156e0a46cc3f5abd6b11051dfb", size = 10909806, upload-time = "2023-11-10T19:13:48.923Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/8e/e4/29e1fa38caba5d5d6db1807ff0f4152b51b7fb3af3a8446095b0a7619d54/pandas-2.1.3-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:35172bff95f598cc5866c047f43c7f4df2c893acd8e10e6653a4b792ed7f19bb", size = 14768086, upload-time = "2023-11-10T19:14:07.281Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/1b/fa/4e5d054549faf1524230ffcd57ca98bb7350a4ed62ef722daabde4cb7632/pandas-2.1.3-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:59dfe0e65a2f3988e940224e2a70932edc964df79f3356e5f2997c7d63e758b4", size = 12263333, upload-time = "2023-11-10T19:14:21.703Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/49/73/5fe14264e2c9e53f40f946782f0e38240f0a6d577609aca440f7223facd6/pandas-2.1.3-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:0296a66200dee556850d99b24c54c7dfa53a3264b1ca6f440e42bad424caea03", size = 13065932, upload-time = "2023-11-10T19:14:40.51Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/3d/c6/9bb3a165e915b9a43b2fd1d35620977bf1371e08538f3649585a1d7b4794/pandas-2.1.3-cp310-cp310-win_amd64.whl", hash = "sha256:465571472267a2d6e00657900afadbe6097c8e1dc43746917db4dfc862e8863e", size = 10666082, upload-time = "2023-11-10T19:14:57.838Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/1b/ef/63136f5ab2dab6f119ded7c5d31b6294e825059b57d4e5b03042fd557b46/pandas-2.1.3-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:04d4c58e1f112a74689da707be31cf689db086949c71828ef5da86727cfe3f82", size = 11618611, upload-time = "2023-11-10T19:15:13.366Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/d6/7c/20e737300f9bec011fb79c01d8948bc38c854876aac2da2cfcdd0992b153/pandas-2.1.3-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:7fa2ad4ff196768ae63a33f8062e6838efed3a319cf938fdf8b95e956c813042", size = 10777582, upload-time = "2023-11-10T19:15:30.139Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/84/d5/dbd0140e9b2cc27566c8213f9f55426487815b3cf1cdabca3edaf5472e35/pandas-2.1.3-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:4441ac94a2a2613e3982e502ccec3bdedefe871e8cea54b8775992485c5660ef", size = 14778000, upload-time = "2023-11-10T19:15:48.391Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/08/de/d4448c423484537ebc9373d3da2496a2e47f42ea11ff48e025cf49665471/pandas-2.1.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:d5ded6ff28abbf0ea7689f251754d3789e1edb0c4d0d91028f0b980598418a58", size = 12231870, upload-time = "2023-11-10T19:16:04.867Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/8c/76/26e0d50e9fa0ddedd23f0098e5940273a984aac7a0a84493e07d502b21e2/pandas-2.1.3-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:fca5680368a5139d4920ae3dc993eb5106d49f814ff24018b64d8850a52c6ed2", size = 13065063, upload-time = "2023-11-10T19:16:22.157Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/97/d8/dc2f6bff06a799a5603c414afc6de39c6351fe34892d50b6a077df3be6ac/pandas-2.1.3-cp311-cp311-win_amd64.whl", hash = "sha256:de21e12bf1511190fc1e9ebc067f14ca09fccfb189a813b38d63211d54832f5f", size = 10612232, upload-time = "2023-11-10T19:16:36.174Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/2b/eb/9c6d267f7f35e3150da9abdc70a39e2a98ece154909b61f3ac939ec38811/pandas-2.1.3-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:a5d53c725832e5f1645e7674989f4c106e4b7249c1d57549023ed5462d73b140", size = 11354802, upload-time = "2023-11-10T19:16:51.56Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/a0/68/265225df9e90ade0c332db4148e9aff8c9bcb4e8dd6c681ec4f512770765/pandas-2.1.3-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:7cf4cf26042476e39394f1f86868d25b265ff787c9b2f0d367280f11afbdee6d", size = 10611390, upload-time = "2023-11-10T19:17:05.783Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/64/06/6a7e7135cfe2173edae5f5d0a08f7b4b4e37c36b66618b2a1cc077e338ce/pandas-2.1.3-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:72c84ec1b1d8e5efcbff5312abe92bfb9d5b558f11e0cf077f5496c4f4a3c99e", size = 14246847, upload-time = "2023-11-10T19:17:22.607Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/ae/a5/5d1deab99008002dfe2c6122352687fd4c2f82688775177729cb0d67556d/pandas-2.1.3-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:1f539e113739a3e0cc15176bf1231a553db0239bfa47a2c870283fd93ba4f683", size = 11660377, upload-time = "2023-11-10T19:17:37.247Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/cd/ff/b425420750328dddfd72448712d29f5b7d873f48162c07e70dd86acc3007/pandas-2.1.3-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:fc77309da3b55732059e484a1efc0897f6149183c522390772d3561f9bf96c00", size = 12371733, upload-time = "2023-11-10T19:17:54.732Z" },
-    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/df/92/a3fa053c74198f9f0224b2c04dc74f41d2e14e30329c082f7a657f9ca4c5/pandas-2.1.3-cp312-cp312-win_amd64.whl", hash = "sha256:08637041279b8981a062899da0ef47828df52a1838204d2b3761fbd3e9fcb549", size = 10497639, upload-time = "2023-11-10T19:18:08.466Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/2e/16/a8eeb70aad84ccbf14076793f90e0031eded63c1899aeae9fdfbf37881f4/pandas-2.3.2-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:52bc29a946304c360561974c6542d1dd628ddafa69134a7131fdfd6a5d7a1a35", size = 11539648, upload-time = "2025-08-21T10:26:36.236Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/47/f1/c5bdaea13bf3708554d93e948b7ea74121ce6e0d59537ca4c4f77731072b/pandas-2.3.2-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:220cc5c35ffaa764dd5bb17cf42df283b5cb7fdf49e10a7b053a06c9cb48ee2b", size = 10786923, upload-time = "2025-08-21T10:26:40.518Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/bb/10/811fa01476d29ffed692e735825516ad0e56d925961819e6126b4ba32147/pandas-2.3.2-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:42c05e15111221384019897df20c6fe893b2f697d03c811ee67ec9e0bb5a3424", size = 11726241, upload-time = "2025-08-21T10:26:43.175Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/c4/6a/40b043b06e08df1ea1b6d20f0e0c2f2c4ec8c4f07d1c92948273d943a50b/pandas-2.3.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:cc03acc273c5515ab69f898df99d9d4f12c4d70dbfc24c3acc6203751d0804cf", size = 12349533, upload-time = "2025-08-21T10:26:46.611Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/e2/ea/2e081a2302e41a9bca7056659fdd2b85ef94923723e41665b42d65afd347/pandas-2.3.2-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:d25c20a03e8870f6339bcf67281b946bd20b86f1a544ebbebb87e66a8d642cba", size = 13202407, upload-time = "2025-08-21T10:26:49.068Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/f4/12/7ff9f6a79e2ee8869dcf70741ef998b97ea20050fe25f83dc759764c1e32/pandas-2.3.2-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:21bb612d148bb5860b7eb2c10faacf1a810799245afd342cf297d7551513fbb6", size = 13837212, upload-time = "2025-08-21T10:26:51.832Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/d8/df/5ab92fcd76455a632b3db34a746e1074d432c0cdbbd28d7cd1daba46a75d/pandas-2.3.2-cp310-cp310-win_amd64.whl", hash = "sha256:b62d586eb25cb8cb70a5746a378fc3194cb7f11ea77170d59f889f5dfe3cec7a", size = 11338099, upload-time = "2025-08-21T10:26:54.382Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/7a/59/f3e010879f118c2d400902d2d871c2226cef29b08c09fb8dc41111730400/pandas-2.3.2-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:1333e9c299adcbb68ee89a9bb568fc3f20f9cbb419f1dd5225071e6cddb2a743", size = 11563308, upload-time = "2025-08-21T10:26:56.656Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/38/18/48f10f1cc5c397af59571d638d211f494dba481f449c19adbd282aa8f4ca/pandas-2.3.2-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:76972bcbd7de8e91ad5f0ca884a9f2c477a2125354af624e022c49e5bd0dfff4", size = 10820319, upload-time = "2025-08-21T10:26:59.162Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/95/3b/1e9b69632898b048e223834cd9702052bcf06b15e1ae716eda3196fb972e/pandas-2.3.2-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:b98bdd7c456a05eef7cd21fd6b29e3ca243591fe531c62be94a2cc987efb5ac2", size = 11790097, upload-time = "2025-08-21T10:27:02.204Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/8b/ef/0e2ffb30b1f7fbc9a588bd01e3c14a0d96854d09a887e15e30cc19961227/pandas-2.3.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:1d81573b3f7db40d020983f78721e9bfc425f411e616ef019a10ebf597aedb2e", size = 12397958, upload-time = "2025-08-21T10:27:05.409Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/23/82/e6b85f0d92e9afb0e7f705a51d1399b79c7380c19687bfbf3d2837743249/pandas-2.3.2-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:e190b738675a73b581736cc8ec71ae113d6c3768d0bd18bffa5b9a0927b0b6ea", size = 13225600, upload-time = "2025-08-21T10:27:07.791Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/e8/f1/f682015893d9ed51611948bd83683670842286a8edd4f68c2c1c3b231eef/pandas-2.3.2-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:c253828cb08f47488d60f43c5fc95114c771bbfff085da54bfc79cb4f9e3a372", size = 13879433, upload-time = "2025-08-21T10:27:10.347Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/a7/e7/ae86261695b6c8a36d6a4c8d5f9b9ede8248510d689a2f379a18354b37d7/pandas-2.3.2-cp311-cp311-win_amd64.whl", hash = "sha256:9467697b8083f9667b212633ad6aa4ab32436dcbaf4cd57325debb0ddef2012f", size = 11336557, upload-time = "2025-08-21T10:27:12.983Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/ec/db/614c20fb7a85a14828edd23f1c02db58a30abf3ce76f38806155d160313c/pandas-2.3.2-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:3fbb977f802156e7a3f829e9d1d5398f6192375a3e2d1a9ee0803e35fe70a2b9", size = 11587652, upload-time = "2025-08-21T10:27:15.888Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/99/b0/756e52f6582cade5e746f19bad0517ff27ba9c73404607c0306585c201b3/pandas-2.3.2-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:1b9b52693123dd234b7c985c68b709b0b009f4521000d0525f2b95c22f15944b", size = 10717686, upload-time = "2025-08-21T10:27:18.486Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/37/4c/dd5ccc1e357abfeee8353123282de17997f90ff67855f86154e5a13b81e5/pandas-2.3.2-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0bd281310d4f412733f319a5bc552f86d62cddc5f51d2e392c8787335c994175", size = 11278722, upload-time = "2025-08-21T10:27:21.149Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/d3/a4/f7edcfa47e0a88cda0be8b068a5bae710bf264f867edfdf7b71584ace362/pandas-2.3.2-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:96d31a6b4354e3b9b8a2c848af75d31da390657e3ac6f30c05c82068b9ed79b9", size = 11987803, upload-time = "2025-08-21T10:27:23.767Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/f6/61/1bce4129f93ab66f1c68b7ed1c12bac6a70b1b56c5dab359c6bbcd480b52/pandas-2.3.2-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:df4df0b9d02bb873a106971bb85d448378ef14b86ba96f035f50bbd3688456b4", size = 12766345, upload-time = "2025-08-21T10:27:26.6Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/8e/46/80d53de70fee835531da3a1dae827a1e76e77a43ad22a8cd0f8142b61587/pandas-2.3.2-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:213a5adf93d020b74327cb2c1b842884dbdd37f895f42dcc2f09d451d949f811", size = 13439314, upload-time = "2025-08-21T10:27:29.213Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/28/30/8114832daff7489f179971dbc1d854109b7f4365a546e3ea75b6516cea95/pandas-2.3.2-cp312-cp312-win_amd64.whl", hash = "sha256:8c13b81a9347eb8c7548f53fd9a4f08d4dfe996836543f805c987bafa03317ae", size = 10983326, upload-time = "2025-08-21T10:27:31.901Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/27/64/a2f7bf678af502e16b472527735d168b22b7824e45a4d7e96a4fbb634b59/pandas-2.3.2-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:0c6ecbac99a354a051ef21c5307601093cb9e0f4b1855984a084bfec9302699e", size = 11531061, upload-time = "2025-08-21T10:27:34.647Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/54/4c/c3d21b2b7769ef2f4c2b9299fcadd601efa6729f1357a8dbce8dd949ed70/pandas-2.3.2-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:c6f048aa0fd080d6a06cc7e7537c09b53be6642d330ac6f54a600c3ace857ee9", size = 10668666, upload-time = "2025-08-21T10:27:37.203Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/50/e2/f775ba76ecfb3424d7f5862620841cf0edb592e9abd2d2a5387d305fe7a8/pandas-2.3.2-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0064187b80a5be6f2f9c9d6bdde29372468751dfa89f4211a3c5871854cfbf7a", size = 11332835, upload-time = "2025-08-21T10:27:40.188Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/8f/52/0634adaace9be2d8cac9ef78f05c47f3a675882e068438b9d7ec7ef0c13f/pandas-2.3.2-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:4ac8c320bded4718b298281339c1a50fb00a6ba78cb2a63521c39bec95b0209b", size = 12057211, upload-time = "2025-08-21T10:27:43.117Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/0b/9d/2df913f14b2deb9c748975fdb2491da1a78773debb25abbc7cbc67c6b549/pandas-2.3.2-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:114c2fe4f4328cf98ce5716d1532f3ab79c5919f95a9cfee81d9140064a2e4d6", size = 12749277, upload-time = "2025-08-21T10:27:45.474Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/87/af/da1a2417026bd14d98c236dba88e39837182459d29dcfcea510b2ac9e8a1/pandas-2.3.2-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:48fa91c4dfb3b2b9bfdb5c24cd3567575f4e13f9636810462ffed8925352be5a", size = 13415256, upload-time = "2025-08-21T10:27:49.885Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/22/3c/f2af1ce8840ef648584a6156489636b5692c162771918aa95707c165ad2b/pandas-2.3.2-cp313-cp313-win_amd64.whl", hash = "sha256:12d039facec710f7ba305786837d0225a3444af7bbd9c15c32ca2d40d157ed8b", size = 10982579, upload-time = "2025-08-21T10:28:08.435Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/f3/98/8df69c4097a6719e357dc249bf437b8efbde808038268e584421696cbddf/pandas-2.3.2-cp313-cp313t-macosx_10_13_x86_64.whl", hash = "sha256:c624b615ce97864eb588779ed4046186f967374185c047070545253a52ab2d57", size = 12028163, upload-time = "2025-08-21T10:27:52.232Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/0e/23/f95cbcbea319f349e10ff90db488b905c6883f03cbabd34f6b03cbc3c044/pandas-2.3.2-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:0cee69d583b9b128823d9514171cabb6861e09409af805b54459bd0c821a35c2", size = 11391860, upload-time = "2025-08-21T10:27:54.673Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/ad/1b/6a984e98c4abee22058aa75bfb8eb90dce58cf8d7296f8bc56c14bc330b0/pandas-2.3.2-cp313-cp313t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:2319656ed81124982900b4c37f0e0c58c015af9a7bbc62342ba5ad07ace82ba9", size = 11309830, upload-time = "2025-08-21T10:27:56.957Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/15/d5/f0486090eb18dd8710bf60afeaf638ba6817047c0c8ae5c6a25598665609/pandas-2.3.2-cp313-cp313t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:b37205ad6f00d52f16b6d09f406434ba928c1a1966e2771006a9033c736d30d2", size = 11883216, upload-time = "2025-08-21T10:27:59.302Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/10/86/692050c119696da19e20245bbd650d8dfca6ceb577da027c3a73c62a047e/pandas-2.3.2-cp313-cp313t-musllinux_1_2_aarch64.whl", hash = "sha256:837248b4fc3a9b83b9c6214699a13f069dc13510a6a6d7f9ba33145d2841a012", size = 12699743, upload-time = "2025-08-21T10:28:02.447Z" },
+    { url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/packages/cd/d7/612123674d7b17cf345aad0a10289b2a384bff404e0463a83c4a3a59d205/pandas-2.3.2-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:d2c3554bd31b731cd6490d94a28f3abb8dd770634a9e06eb6d2911b9827db370", size = 13186141, upload-time = "2025-08-21T10:28:05.377Z" },
 ]
 
 [[package]]
diff --git a/webui.py b/webui.py
index ecac7f2..13fecf8 100644
--- a/webui.py
+++ b/webui.py
@@ -7,11 +7,11 @@ import time
 
 import warnings
 
-import pandas as pd
-
 warnings.filterwarnings("ignore", category=FutureWarning)
 warnings.filterwarnings("ignore", category=UserWarning)
 
+import pandas as pd
+
 current_dir = os.path.dirname(os.path.abspath(__file__))
 sys.path.append(current_dir)
 sys.path.append(os.path.join(current_dir, "indextts"))
@@ -209,9 +209,9 @@ with gr.Blocks(title="IndexTTS Demo") as demo:
         with gr.Accordion(i18n("高级生成参数设置"), open=False):
             with gr.Row():
                 with gr.Column(scale=1):
-                    gr.Markdown(f"**{i18n('GPT2 采样设置')}** _{i18n('参数会影响音频多样性和生成速度详见')}[Generation strategies](https://huggingface.co/docs/transformers/main/en/generation_strategies)_")
+                    gr.Markdown(f"**{i18n('GPT2 采样设置')}** _{i18n('参数会影响音频多样性和生成速度详见')} [Generation strategies](https://huggingface.co/docs/transformers/main/en/generation_strategies)._")
                     with gr.Row():
-                        do_sample = gr.Checkbox(label="do_sample", value=True, info="是否进行采样")
+                        do_sample = gr.Checkbox(label="do_sample", value=True, info=i18n("是否进行采样"))
                         temperature = gr.Slider(label="temperature", minimum=0.1, maximum=2.0, value=0.8, step=0.1)
                     with gr.Row():
                         top_p = gr.Slider(label="top_p", minimum=0.0, maximum=1.0, value=0.8, step=0.01)
@@ -220,7 +220,7 @@ with gr.Blocks(title="IndexTTS Demo") as demo:
                     with gr.Row():
                         repetition_penalty = gr.Number(label="repetition_penalty", precision=None, value=10.0, minimum=0.1, maximum=20.0, step=0.1)
                         length_penalty = gr.Number(label="length_penalty", precision=None, value=0.0, minimum=-2.0, maximum=2.0, step=0.1)
-                    max_mel_tokens = gr.Slider(label="max_mel_tokens", value=1500, minimum=50, maximum=tts.cfg.gpt.max_mel_tokens, step=10, info="生成Token最大数量，过小导致音频被截断", key="max_mel_tokens")
+                    max_mel_tokens = gr.Slider(label="max_mel_tokens", value=1500, minimum=50, maximum=tts.cfg.gpt.max_mel_tokens, step=10, info=i18n("生成Token最大数量，过小导致音频被截断"), key="max_mel_tokens")
                     # with gr.Row():
                     #     typical_sampling = gr.Checkbox(label="typical_sampling", value=False, info="不建议使用")
                     #     typical_mass = gr.Slider(label="typical_mass", value=0.9, minimum=0.0, maximum=1.0, step=0.1)