Update app.py

app.py

@@ -130,22 +130,27 @@ def infer(language,text,speaker_id, n_scale= 0.667,n_scale_w = 0.8, l_scale = 1
     final_list = extrac(text.replace('“','').replace('”',''))
     audio_fin = []
     c = 0
+    t = datetime.timedelta(seconds=0)
+    f1 = open("lovelive/subtitles.srt",'w',encoding='utf-8')
     for sentence in final_list:
         c +=1
-        try:
-            stn_tst = get_text(sle(language,sentence), hps_ms)
-            with torch.no_grad():
-                x_tst = stn_tst.unsqueeze(0).to(dev)
-                x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).to(dev)
-                sid = torch.LongTensor([speaker_id]).to(dev)
-                t1 = time.time()
-                audio = net_g_ms.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=n_scale, noise_scale_w=n_scale_w, length_scale=l_scale)[0][0,0].data.cpu().float().numpy()
-                t2 = time.time()
-                spending_time = "第"+str(c)+"句的推理时间为："+str(t2-t1)+"s"
-                print(spending_time)
-                audio_fin.append(audio)
-        except:
-            print('存在非法字符')
+        stn_tst = get_text(sle(language,sentence), hps_ms)
+        with torch.no_grad():
+            x_tst = stn_tst.unsqueeze(0).to(dev)
+            x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).to(dev)
+            sid = torch.LongTensor([speaker_id]).to(dev)
+            t1 = time.time()
+            audio = net_g_ms.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=n_scale, noise_scale_w=n_scale_w, length_scale=l_scale)[0][0,0].data.cpu().float().numpy()
+            t2 = time.time()
+            spending_time = "第"+str(c)+"句的推理时间为："+str(t2-t1)+"s"
+            print(spending_time)
+            time_start = str(t).split(".")[0] + "," + str(t.microseconds)[:3]
+            last_time = datetime.timedelta(seconds=len(audio)/float(22050))
+            t+=last_time
+            time_end = str(t).split(".")[0] + "," + str(t.microseconds)[:3]
+            print(time_end)
+            f1.write(str(c-1)+'\n'+time_start+' --> '+time_end+'\n'+sentence+'\n\n')
+            audio_fin.append(audio)
     return (hps_ms.data.sampling_rate, np.concatenate(audio_fin))
 lan = ["中文","日文","自动"]
 idols = ["高咲侑","歩夢","かすみ","しずく","果林","愛","せつ菜","璃奈","栞子","エマ","ランジュ","ミア","派蒙"]
@@ -171,7 +176,7 @@ with app:
             para_input3 = gr.Slider(minimum= 0.1,maximum=10,label="更改时间比例", value=1)
             tts_submit = gr.Button("Generate", variant="primary")
             speaker1 = gr.Dropdown(label="选择说话人",choices=idols, value="歩夢", interactive=True)
-            tts_output2 = gr.Audio(label="Output")
+            tts_output2 = gr.Audio(label="字幕见 https://huggingface.co/spaces/Mahiruoshi/Lovelive_Nijigasaki_VITS/lovelive/subtitles.srt")
             tts_submit.click(infer, [language,tts_input1,speaker1,para_input1,para_input2,para_input3], [tts_output2])
     #app.launch(share=True)
     app.launch()