diff --git a/GPT_SoVITS/VoiceSave/__init__.py b/GPT_SoVITS/VoiceSave/__init__.py
new file mode 100644
index 00000000..dadc19e7
--- /dev/null
+++ b/GPT_SoVITS/VoiceSave/__init__.py
@@ -0,0 +1,143 @@
+import zipfile
+from . import file_lib as fl
+from . import time_lib as tl
+from . import info_lib as il
+import os
+from typing import Union
+import numpy as np
+import torch
+
+POOL:set = set() 
+def get_unique_name(name,MySet:set=set()):
+    _id = 1
+    if name not in POOL and name not in MySet:
+        POOL.add(name)
+        return name
+    while name in POOL or name in MySet:
+        _id += 1
+        name = f'{name}_{_id}'
+    POOL.add(name)
+    return name
+
+TEMP_DIR = fl.merge_dir_txt2(fl.get_my_dir(), "Temp")
+TEMP_ZIP_DIR = fl.merge_dir_txt2(TEMP_DIR, "ZipTemp")
+def _tensor_to_numpy(tensor: torch.Tensor) -> np.ndarray:
+    cloned = tensor.clone().detach()
+    np_array = cloned.cpu().numpy()
+    return np_array
+
+def save_np(path: str, np_array: np.ndarray) -> None:
+    np.save(path, np_array)
+
+class ZIP_File:
+    def __init__(self, path: str,name:str,MySet:set=set()):
+        self.path = path
+        if not os.path.exists(self.path):
+            with zipfile.ZipFile(self.path, 'w') as zipf:
+                pass
+        self.name = get_unique_name(name,MySet=MySet)#MySet用于补充命名集合，防止文件夹混淆
+        self.temp_write = fl.merge_dir_txt2(TEMP_ZIP_DIR, self.name)
+
+        if not os.path.exists(self.temp_write):
+            os.makedirs(self.temp_write)
+
+    def release(self):
+        '''relaese the zip file, extract it to temp dir'''
+        if os.path.exists(self.temp_write):
+            fl.delete_dir(self.temp_write)
+            fl.create_dir(self.temp_write)
+        with zipfile.ZipFile(self.path, 'r') as zipf:
+            zipf.extractall(self.temp_write)
+        #fl.delete_file(self.path)
+    def create_dir(self, dir_:str):
+        dir_path = fl.merge_dir_txt2(self.temp_write, dir_)
+        if not os.path.exists(dir_path):
+            os.makedirs(dir_path,exist_ok=True)
+
+    def create_file(self, file_name:str,location:str=''):
+        if location == '':
+            file_path = fl.merge_dir_txt2(self.temp_write,file_name)
+        else:
+            file_path = fl.merge_dir_txt2(self.temp_write, location, file_name)
+        if not os.path.exists(file_path):
+            os.makedirs(os.path.dirname(file_path),exist_ok=True)
+        with open(file_path, 'w') as f:
+            pass
+
+    def get_file_path(self, file_name:str,location:str=''):
+        if location == '':
+            file_path = fl.merge_dir_txt2(self.temp_write,file_name)
+        else:
+            file_path = fl.merge_dir_txt2(self.temp_write, location, file_name)
+        if not os.path.exists(file_path):
+            raise FileNotFoundError(f"File {file_path} does not exist.")
+        return file_path
+    
+    def get_file_obj(self, file_name:str,location:str='',mode:str='r'):
+        if location == '':
+            file_path = fl.merge_dir_txt2(self.temp_write,file_name)
+        else:
+            file_path = fl.merge_dir_txt2(self.temp_write, location, file_name)
+        if not os.path.exists(file_path):
+            raise FileNotFoundError(f"File {file_path} does not exist.")
+        return open(file_path, mode)
+    
+    def save_file(self, obj):
+        obj.close()
+    
+    def save_zip(self):
+        with zipfile.ZipFile(self.path, 'w', zipfile.ZIP_DEFLATED) as zipf:
+            for root, dirs, files in os.walk(self.temp_write):
+                for file in files:
+                    file_path = os.path.join(root, file)
+                    arcname = os.path.relpath(file_path, self.temp_write)
+                    zipf.write(file_path, arcname)
+        #fl.delete_dir(self.temp_write)
+
+    def close(self):
+        self.save_zip()
+        fl.delete_dir(self.temp_write)
+        POOL.remove(self.name)
+
+def save_tensor(path: str, tensors: Union[torch.Tensor, list],name:str,MySet:set=set(),file_names:Union[str,list,None]=None,**info_save) -> None:
+    if isinstance(tensors, torch.Tensor):
+        tensors = [tensors]
+    if not file_names:
+        return
+    if isinstance(file_names, str):
+        files = [file_names]
+    else:
+        files = file_names
+
+    if len(tensors) != len(files):
+        raise ValueError("The number of tensors and files must be the same.")
+    np_arrays = []
+    for tensor in tensors:
+        np_array = _tensor_to_numpy(tensor)
+        np_arrays.append(np_array)
+    zf = ZIP_File(path, name, MySet=MySet)
+    zf.create_file("voice.json")
+    info = {'name': name}
+    info.update(info_save)
+    il.save_info(info, str(zf.get_file_path("voice.json")))
+    for i in range(len(files)):
+        file_name = files[i]
+        np_array = np_arrays[i]
+        zf.create_file(file_name)
+        save_np(str(zf.get_file_path(file_name)), np_array)
+    zf.close()
+    del zf
+
+def load_tensor(path: str,name:str,find_func,MySet:set=set()) -> list[torch.Tensor]:
+    zf = ZIP_File(path, name, MySet=MySet)
+    zf.release()
+    voice_path = find_func(zf,il)
+    tensors = []
+    for i in range(len(voice_path)):
+        v = voice_path[i]
+        np_array = np.load(v,allow_pickle=True)
+        tensor = torch.from_numpy(np_array)
+        tensors.append(tensor)
+    zf.close()
+    del zf
+    return tensors
\ No newline at end of file
diff --git a/GPT_SoVITS/VoiceSave/file_lib.py b/GPT_SoVITS/VoiceSave/file_lib.py
new file mode 100644
index 00000000..e26b7c5a
--- /dev/null
+++ b/GPT_SoVITS/VoiceSave/file_lib.py
@@ -0,0 +1,35 @@
+import os
+import shutil
+from pathlib import Path
+
+def get_my_dir():
+    return os.path.dirname(os.path.abspath(__file__))
+
+def get_parent_dir(dir_path,depth=1):
+    parent_path = Path(dir_path)
+    for _ in range(depth):
+        parent_path = parent_path.parent
+    return parent_path
+
+def merge_dir_txt(a,b):
+    c=os.path.join(a,b)
+    return c
+def merge_dir_txt2(*TXT):
+    return Path(os.path.join(*TXT))
+def create_dir(path: Path, overwrite=False) -> bool:
+    if overwrite and path.exists():
+        shutil.rmtree(path)
+    path = Path(path)
+    path.mkdir(parents=True, exist_ok=True)
+    return path.exists()
+def get_dir_children_dirs(path: Path):
+    return [item for item in path.iterdir() if item.is_dir()]
+def get_dir_children_files(path: Path):
+    return [item for item in path.iterdir() if item.is_file()]
+def delete_dir(path: Path):
+    return shutil.rmtree(path)
+def delete_file(path: Path):
+    return os.remove(path)
+def file_exists(path: Path):
+    path = Path(path)
+    return path.exists()
\ No newline at end of file
diff --git a/GPT_SoVITS/VoiceSave/info_lib.py b/GPT_SoVITS/VoiceSave/info_lib.py
new file mode 100644
index 00000000..c8f825b0
--- /dev/null
+++ b/GPT_SoVITS/VoiceSave/info_lib.py
@@ -0,0 +1,10 @@
+import json
+
+def load_info(info_path):
+    with open(info_path, 'r', encoding='utf-8') as f:
+        info = json.load(f)
+    return info
+
+def save_info(info, info_path):
+    with open(info_path, 'w', encoding='utf-8') as f:
+        json.dump(info, f, ensure_ascii=False, indent=4)
\ No newline at end of file
diff --git a/GPT_SoVITS/VoiceSave/time_lib.py b/GPT_SoVITS/VoiceSave/time_lib.py
new file mode 100644
index 00000000..819f90f1
--- /dev/null
+++ b/GPT_SoVITS/VoiceSave/time_lib.py
@@ -0,0 +1,38 @@
+import time
+#time styles
+STYLE_Y = "%Y"
+STYLE_M = "%m"
+STYLE_D = "%d"
+STYLE_H = "%H"
+STYLE_MIN = "%M"
+STYLE_S = "%S"
+STYLE_FULL = "%Y-%m-%d_%H.%M.%S"
+#quick calls
+def get_time_y(STYLE = STYLE_Y):
+    return time.strftime(STYLE, time.localtime())
+def get_time_m(STYLE = STYLE_M):
+    return time.strftime(STYLE, time.localtime())
+def get_time_d(STYLE = STYLE_D):
+    return time.strftime(STYLE, time.localtime())
+def get_time_h(STYLE = STYLE_H):
+    return time.strftime(STYLE, time.localtime())
+def get_time_min(STYLE = STYLE_MIN):
+    return time.strftime(STYLE, time.localtime())
+def get_time_s(STYLE = STYLE_S):
+    return time.strftime(STYLE, time.localtime())
+def get_time_full(STYLE = STYLE_FULL):
+    return time.strftime(STYLE, time.localtime())
+
+def s(t:float):
+    time.sleep(t)
+    return 
+###
+
+if __name__ == '__main__':
+    print(get_time_y())
+    print(get_time_m())
+    print(get_time_d())
+    print(get_time_h())
+    print(get_time_min())
+    print(get_time_s())
+    print(get_time_full())
\ No newline at end of file
diff --git a/GPT_SoVITS/inference_webui.py b/GPT_SoVITS/inference_webui.py
index a361ed58..3031b9ba 100644
--- a/GPT_SoVITS/inference_webui.py
+++ b/GPT_SoVITS/inference_webui.py
@@ -8,6 +8,51 @@
 """
 import psutil
 import os
+import sys
+from pathlib import Path
+
+def get_my_dir():
+    return os.path.dirname(os.path.abspath(__file__))
+
+def get_parent_dir(dir_path,depth=1):
+    parent_path = Path(dir_path)
+    for _ in range(depth):
+        parent_path = parent_path.parent
+    return parent_path
+
+def merge_dir_txt2(*TXT):
+    return Path(os.path.join(*TXT))
+
+ROOT_DIR = str(get_parent_dir(get_my_dir()))
+sys.path.append(get_my_dir())
+import VoiceSave
+
+POOL:set = set() 
+def _get_unique_name(name,MySet:set=set()):
+    _id = 1
+    if name not in POOL and name not in MySet:
+        POOL.add(name)
+        return name
+    while name in POOL or name in MySet:
+        _id += 1
+        name = f'{name}_{_id}'
+    POOL.add(name)
+    return name
+
+def find_func(zf,il):
+    f = zf.get_file_path("voice.json")
+    info = il.load_info(f)
+    if info is None:
+        return None
+    list_names = info["access_list"]
+    ret = []
+    for name in list_names:
+        try:
+            a = zf.get_file_path(name)
+            ret.append(a)
+        except FileNotFoundError:
+            continue
+    return ret
 
 def set_high_priority():
     """把当前 Python 进程设为 HIGH_PRIORITY_CLASS"""
@@ -765,6 +810,18 @@ def get_tts_wav(
     sample_steps=8,
     if_sr=False,
     pause_second=0.3,
+
+    SaveSvEmb=False,
+    SaveRefers=False,
+    SaveSvEmbName="sv_emb.voice",
+    SaveRefersName="refers.voice",
+
+    InjectSvEmb=False,
+    InjectRefers=False,
+    InjectSvEmbName="sv_emb.voice",
+    InjectRefersName="refers.voice",
+
+    EnableAudioLoad=True,
 ):
     global cache
     if ref_wav_path:
@@ -898,20 +955,92 @@ def get_tts_wav(
                 sv_emb = []
                 if sv_cn_model == None:
                     init_sv_cn()
-            if inp_refs:
-                for path in inp_refs:
-                    try:  #####这里加上提取sv的逻辑，要么一堆sv一堆refer，要么单个sv单个refer
-                        refer, audio_tensor = get_spepc(hps, path.name, dtype, device, is_v2pro)
-                        refers.append(refer)
+
+            try:
+                if EnableAudioLoad:
+                    if inp_refs:
+                        for path in inp_refs:
+                            try:  #####这里加上提取sv的逻辑，要么一堆sv一堆refer，要么单个sv单个refer
+                                refer, audio_tensor = get_spepc(hps, path.name, dtype, device, is_v2pro)
+                                refers.append(refer)
+                                if is_v2pro:
+                                    sv_emb.append(sv_cn_model.compute_embedding3(audio_tensor))
+                                #print("refer:", refer.shape)
+                            except:
+                                traceback.print_exc()
+                    if len(refers) == 0:
+                        refers, audio_tensor = get_spepc(hps, ref_wav_path, dtype, device, is_v2pro)
+                        refers = [refers]
                         if is_v2pro:
-                            sv_emb.append(sv_cn_model.compute_embedding3(audio_tensor))
-                    except:
-                        traceback.print_exc()
-            if len(refers) == 0:
-                refers, audio_tensor = get_spepc(hps, ref_wav_path, dtype, device, is_v2pro)
-                refers = [refers]
-                if is_v2pro:
-                    sv_emb = [sv_cn_model.compute_embedding3(audio_tensor)]
+                            sv_emb = [sv_cn_model.compute_embedding3(audio_tensor)]
+                else:
+                    refers = []
+                    sv_emb = []
+            except:
+                traceback.print_exc()
+
+            try:
+                if SaveSvEmb and is_v2pro:
+                    names = []
+                    for i in sv_emb:
+                        names.append(_get_unique_name(str(i.shape))+".npy")
+                    sv_path = merge_dir_txt2(ROOT_DIR,"output","sv_emb_opt")
+                    if not os.path.exists(sv_path):
+                        os.makedirs(sv_path,exist_ok=True)
+                    if not os.path.exists(SaveSvEmbName):
+                        _pth_ = str(merge_dir_txt2(ROOT_DIR,"output","sv_emb_opt",SaveSvEmbName))
+                    else:
+                        _pth_ = SaveSvEmbName
+                    VoiceSave.save_tensor(_pth_,sv_emb,SaveSvEmbName,file_names=names,access_list=names)
+            except:
+                traceback.print_exc()
+
+            try:                
+                if SaveRefers:
+                    names = []
+                    for i in refers:
+                        names.append(_get_unique_name(str(i.shape))+".npy")
+                    refers_path = merge_dir_txt2(ROOT_DIR,"output","refers_opt")
+                    if not os.path.exists(refers_path):
+                        os.makedirs(refers_path,exist_ok=True)
+                    if not os.path.exists(SaveRefersName):
+                        _pth_ = str(merge_dir_txt2(ROOT_DIR,"output","refers_opt",SaveRefersName))
+                    else:
+                        _pth_ = SaveRefersName
+                    VoiceSave.save_tensor(_pth_,refers,SaveRefersName,file_names=names,access_list=names)
+            except:
+                traceback.print_exc()
+            
+            #print("refers数量:", len(refers))
+            #print("sv_emb数量:", len(sv_emb) if is_v2pro else "无sv_emb")
+
+            try:
+                if InjectSvEmb and is_v2pro:
+                    if not os.path.exists(InjectSvEmbName):
+                        _pth_ = str(merge_dir_txt2(ROOT_DIR,"output","sv_emb_opt",InjectSvEmbName))
+                    else:
+                        _pth_ = InjectSvEmbName
+                    _sv_emb = VoiceSave.load_tensor(_pth_,InjectSvEmbName,find_func)
+                    for i in range(len(_sv_emb)):
+                        sv_emb.append(_sv_emb[i].to(device))
+            except:
+                traceback.print_exc()
+
+            try:
+                if InjectRefers:
+                    if not os.path.exists(InjectRefersName):
+                        _pth_ = str(merge_dir_txt2(ROOT_DIR,"output","refers_opt",InjectRefersName))
+                    else:
+                        _pth_ = InjectRefersName
+                    _refers = VoiceSave.load_tensor(_pth_,InjectRefersName,find_func)
+                    for i in range(len(_refers)):
+                        refers.append(_refers[i].to(device))
+            except:
+                traceback.print_exc()
+
+            #print("注入后refers数量:", len(refers))
+            #print("注入后sv_emb数量:", len(sv_emb) if is_v2pro else "无sv_emb")
+
             if is_v2pro:
                 audio = vq_model.decode(
                     pred_semantic, torch.LongTensor(phones2).to(device).unsqueeze(0), refers, speed=speed, sv_emb=sv_emb
diff --git a/config.json b/config.json
new file mode 100644
index 00000000..09a5d85e
--- /dev/null
+++ b/config.json
@@ -0,0 +1,5 @@
+{
+    "GPU_CHECK":{
+        "DisableGPUMemCheck":false
+    }
+}
\ No newline at end of file
diff --git a/config.py b/config.py
index fdc11c0a..76965027 100644
--- a/config.py
+++ b/config.py
@@ -1,11 +1,20 @@
 import os
 import re
 import sys
+import json
+from pathlib import Path
 
 import torch
 
 from tools.i18n.i18n import I18nAuto
 
+current_dir = str(Path(__file__).parent)
+def merge_dir_txt2(*TXT):
+    return Path(os.path.join(*TXT))
+config_json_location = merge_dir_txt2(current_dir,"config.json")
+with open(str(config_json_location),"r") as f:
+    __info__ = f.read()
+
 i18n = I18nAuto(language=os.environ.get("language", "Auto"))
 
 
@@ -159,8 +168,9 @@ def get_device_dtype_sm(idx: int) -> tuple[torch.device, torch.dtype, float, flo
     major, minor = capability
     sm_version = major + minor / 10.0
     is_16_series = bool(re.search(r"16\d{2}", name)) and sm_version == 7.5
-    if mem_gb < 4 or sm_version < 5.3:
-        return cpu, torch.float32, 0.0, 0.0
+    if not __info__["GPU_CHECK"]["DisableGPUMemCheck"]:
+        if mem_gb < 4 or sm_version < 5.3:
+            return cpu, torch.float32, 0.0, 0.0
     if sm_version == 6.1 or is_16_series == True:
         return cuda, torch.float32, sm_version, mem_gb
     if sm_version > 6.1: