diff --git a/GPT_SoVITS/inference_webui.py b/GPT_SoVITS/inference_webui.py
index 42b7fda..ecbbd2f 100644
--- a/GPT_SoVITS/inference_webui.py
+++ b/GPT_SoVITS/inference_webui.py
@@ -314,7 +314,7 @@ def merge_short_text_in_array(texts, threshold):
 
 ##ref_wav_path+prompt_text+prompt_language+text(单个)+text_language+top_k+top_p+temperature
 # cache_tokens={}#暂未实现清理机制
-cache=None
+cache= {}
 def get_tts_wav(ref_wav_path, prompt_text, prompt_language, text, text_language, how_to_cut=i18n("不切"), top_k=20, top_p=0.6, temperature=0.6, ref_free = False,speed=1,if_freeze=False):
     global cache
     if prompt_text is None or len(prompt_text) == 0:
@@ -380,7 +380,7 @@ def get_tts_wav(ref_wav_path, prompt_text, prompt_language, text, text_language,
     if not ref_free:
         phones1,bert1,norm_text1=get_phones_and_bert(prompt_text, prompt_language)
 
-    for text in texts:
+    for i_text,text in enumerate(texts):
         # 解决输入目标文本的空行导致报错的问题
         if (len(text.strip()) == 0):
             continue
@@ -400,7 +400,8 @@ def get_tts_wav(ref_wav_path, prompt_text, prompt_language, text, text_language,
 
         t2 = ttime()
         # cache_key="%s-%s-%s-%s-%s-%s-%s-%s"%(ref_wav_path,prompt_text,prompt_language,text,text_language,top_k,top_p,temperature)
-        if(type(cache)!=type(None)and if_freeze==True):pred_semantic=cache
+        # print(cache.keys(),if_freeze)
+        if(i_text in cache and if_freeze==True):pred_semantic=cache[i_text]
         else:
             with torch.no_grad():
                 pred_semantic, idx = t2s_model.model.infer_panel(
@@ -415,7 +416,7 @@ def get_tts_wav(ref_wav_path, prompt_text, prompt_language, text, text_language,
                     early_stop_num=hz * max_sec,
                 )
                 pred_semantic = pred_semantic[:, -idx:].unsqueeze(0)
-                cache=pred_semantic
+                cache[i_text]=pred_semantic
         t3 = ttime()
         refer = get_spepc(hps, ref_wav_path)  # .to(device)
         if is_half == True: