Update app.py

app.py

@@ -61,38 +61,6 @@ def to_numpy(tensor: torch.Tensor):
     return tensor.detach().cpu().numpy() if tensor.requires_grad \
         else tensor.detach().numpy()
 
-def to_html(chat_history):
-    chat_html = ""
-    for item in chat_history:
-        if item['role'] == 'user':
-            chat_html += f"""
-                <div style="margin-bottom: 20px;">
-                    <div style="text-align: right; margin-right: 20px;">
-                        <span style="background-color: #4CAF50; color: black; padding: 10px; border-radius: 10px; display: inline-block; max-width: 80%; word-wrap: break-word;">
-                            {item['content']}
-                        </span>
-                    </div>
-                </div>
-            """
-        else:
-            chat_html += f"""
-                <div style="margin-bottom: 20px;">
-                    <div style="text-align: left; margin-left: 20px;">
-                        <span style="background-color: white; color: black; padding: 10px; border-radius: 10px; display: inline-block; max-width: 80%; word-wrap: break-word;">
-                            {item['content']}
-                        </span>
-                    </div>
-                </div>
-            """
-    output_html = f"""
-        <div style="height: 400px; overflow-y: scroll; padding: 10px;">
-            {chat_html}
-        </div>
-    """
-    return output_html
-
-
-
 def chatgpt(text):
     messages = []
     try:
@@ -108,7 +76,7 @@ def chatgpt(text):
                 del messages[-2:]
             with open('log.pickle', 'wb') as f:
                 pickle.dump(messages, f)
-            return reply,messages
+            return reply
     except:
         messages.append({"role": "user", "content": text},)
         chat = openai.ChatCompletion.create(model="gpt-3.5-turbo", messages=messages)
@@ -120,7 +88,7 @@ def chatgpt(text):
             del messages[-2:]
         with open('log.pickle', 'wb') as f:
             pickle.dump(messages, f)
-        return reply,messages
+        return reply
 
 def get_symbols_from_json(path):
     assert os.path.isfile(path)
@@ -154,14 +122,12 @@ def get_text(text,hps_ms):
 
 def create_tts_fn(net_g,hps,speaker_id):
     speaker_id = int(speaker_id)
-    def tts_fn(is_gpt,api_key,is_audio,audiopath,repeat_time,text, language, extract, n_scale= 0.667,n_scale_w = 0.8, l_scale = 1 ):
+    def tts_fn(history,is_gpt,api_key,is_audio,audiopath,repeat_time,text, language, extract, n_scale= 0.667,n_scale_w = 0.8, l_scale = 1 ):
         repeat_time = int(repeat_time)
         if is_gpt:
             openai.api_key = api_key
-            text,messages = chatgpt(text)
-            htm = to_html(messages)
-        else:
-             htm = ""
+            text = chatgpt(text)
+            history[-1][1] = text
         if not extract:
             print(text)
             t1 = time.time()
@@ -175,6 +141,7 @@ def create_tts_fn(net_g,hps,speaker_id):
                 spending_time = "推理时间为："+str(t2-t1)+"s"
                 print(spending_time)
                 file_path = "subtitles.srt"
+                write('moe/temp.wav',22050,audio)
             try:
                 write(audiopath + '.wav',22050,audio)
                 if is_audio:
@@ -183,7 +150,7 @@ def create_tts_fn(net_g,hps,speaker_id):
                         os.system(cmd)
             except:
                 pass
-            return (hps.data.sampling_rate, audio),file_path,htm
+            return history,file_path,(hps.data.sampling_rate,audio)
         else:
             a = ['【','[','(','（']
             b = ['】',']',')','）']
@@ -215,19 +182,22 @@ def create_tts_fn(net_g,hps,speaker_id):
                     print(time_end)
                     f1.write(str(c-1)+'\n'+time_start+' --> '+time_end+'\n'+sentence+'\n\n')
                     audio_fin.append(audio)
-            try:
-                write(audiopath + '.wav',22050,np.concatenate(audio_fin))
-                if is_audio:
-                    for i in range(repeat_time):
-                        cmd = 'ffmpeg -y -i ' +  audiopath + '.wav' + ' -ar 44100 '+ audiopath.replace('temp','temp'+str(i))
-                        os.system(cmd)
+                try:
+                    write(audiopath + '.wav',22050,np.concatenate(audio_fin))
+                    if is_audio:
+                        for i in range(repeat_time):
+                            cmd = 'ffmpeg -y -i ' +  audiopath + '.wav' + ' -ar 44100 '+ audiopath.replace('temp','temp'+str(i))
+                            os.system(cmd)
+                    
+                except:
+                    pass
                 
-            except:
-                pass
             file_path = "subtitles.srt"
-            return (hps.data.sampling_rate, np.concatenate(audio_fin)),file_path,htm
+            return history,file_path,(hps.data.sampling_rate, np.concatenate(audio_fin))
     return tts_fn
 
+def bot(history,user_message):
+    return history + [[user_message, None]]
 
 if __name__ == '__main__':
     hps = utils.get_hparams_from_file('checkpoints/tmp/config.json')
@@ -275,13 +245,13 @@ if __name__ == '__main__':
                                             f'<img style="width:auto;height:400px;" src="file/image/{name}.png">' 
                                             '</div>'
                                         )
-                                    with gr.Accordion(label="聊天记录", open=False):
-                                        with gr.Row(): 
-                                            output_UI = gr.outputs.HTML()
-                                    with gr.Column():
-                                        input1 = gr.TextArea(label="Text", value=example, lines=3)    
-                                        btnVC = gr.Button("Submit")
-                                        output1 = gr.Audio(label="采样率22050")
+                                    chatbot = gr.Chatbot(elem_id="History")
+                                with gr.Row():
+                                    with gr.Column(scale=0.85):
+                                        input1 = gr.TextArea(label="Text", value=example,lines = 1)    
+                                    with gr.Column(scale=0.15, min_width=0):
+                                        btnVC = gr.Button("Send")
+                                output1 = gr.Audio(label="采样率22050")
                                 with gr.Accordion(label="Setting", open=False):
                                     input2 = gr.Dropdown(label="Language", choices=lan, value="自动", interactive=True)
                                     input3 = gr.Checkbox(value=False, label="长句切割(小说合成)")
@@ -291,12 +261,13 @@ if __name__ == '__main__':
                                 with gr.Accordion(label="Advanced Setting", open=False):
                                     audio_input3 = gr.Dropdown(label="重复次数", choices=list(range(101)), value='0', interactive=True) 
                                     api_input1 = gr.Checkbox(value=False, label="接入chatgpt")
-                                    api_input2 = gr.TextArea(label="api-key",lines=1,value = '去官网找')   
+                                    api_input2 = gr.TextArea(label="api-key",lines=1,value = '见 https://openai.com/blog/openai-api')   
                                     output2 = gr.outputs.File(label="字幕文件：subtitles.srt")
                                     audio_input1 = gr.Checkbox(value=False, label="修改音频路径(live2d)")
-                                    audio_input2 = gr.TextArea(label="音频路径",lines=1,value = '示范：D:/app_develop/live2d_whole/2010002/sounds/temp.wav')
-                         
-                                                      
-                        btnVC.click(tts_fn, inputs=[api_input1,api_input2,audio_input1,audio_input2,audio_input3,input1,input2,input3,input4,input5,input6], outputs=[output1,output2,output_UI])
+                                    audio_input2 = gr.TextArea(label="音频路径",lines=1,value = '#参考 D:/app_develop/live2d_whole/2010002/sounds/temp.wav')
+                                 
+                        btnVC.click(bot, inputs = [chatbot,input1], outputs = [chatbot]).then(
+    tts_fn, inputs=[chatbot,api_input1,api_input2,audio_input1,audio_input2,audio_input3,input1,input2,input3,input4,input5,input6], outputs=[chatbot,output2,output1]
+    )
                 
-    app.launch()
+    app.launch()