Update app.py

app.py

@@ -24,7 +24,18 @@ from text import cleaned_text_to_sequence, get_bert
 from text.cleaner import clean_text
 import gradio as gr
 import webbrowser
+import openai
 
+# openai.log = "debug"
+openai.api_base = "https://api.chatanywhere.com.cn/v1"
+
+
+# 非流式响应
+
+def gpt_35_api(gptkey, message):
+    openai.api_key = "sk-" + gptkey
+    completion = openai.ChatCompletion.create(model="gpt-3.5-turbo", messages=[{"role": "user", "content": message}])
+    return completion.choices[0].message.content
 
 net_g = None
 
@@ -51,9 +62,11 @@ def get_text(text, language_str, hps):
 
     return bert, phone, tone, language
 
-def infer(text, sdp_ratio, noise_scale, noise_scale_w, length_scale, sid):
+def infer(text, key, sdp_ratio, noise_scale, noise_scale_w, length_scale, sid):
     global net_g
-    bert, phones, tones, lang_ids = get_text(text, "ZH", hps)
+    message = gpt_35_api(key, text)
+    print(message)
+    bert, phones, tones, lang_ids = get_text(message, "ZH", hps)
     with torch.no_grad():
         x_tst=phones.to(device).unsqueeze(0)
         tones=tones.to(device).unsqueeze(0)
@@ -67,15 +80,15 @@ def infer(text, sdp_ratio, noise_scale, noise_scale_w, length_scale, sid):
         del x_tst, tones, lang_ids, bert, x_tst_lengths, speakers
         return audio
 
-def tts_fn(text, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale):
+def tts_fn(text, key, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale):
     with torch.no_grad():
-        audio = infer(text, sdp_ratio=sdp_ratio, noise_scale=noise_scale, noise_scale_w=noise_scale_w, length_scale=length_scale, sid=speaker)
+        audio = infer(text, key,sdp_ratio=sdp_ratio, noise_scale=noise_scale, noise_scale_w=noise_scale_w, length_scale=length_scale, sid=speaker)
     return "Success", (hps.data.sampling_rate, audio)
 
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
-    parser.add_argument("--model_dir", default="./logs/Ava/G_14300.pth", help="path of your model")
+    parser.add_argument("--model_dir", default="./logs/Bella/G_3100.pth", help="path of your model")
     parser.add_argument("--config_dir", default="./configs/config.json", help="path of your config file")
     parser.add_argument("--share", default=False, help="make link public")
     parser.add_argument("-d", "--debug", action="store_true", help="enable DEBUG-LEVEL log")
@@ -113,40 +126,36 @@ if __name__ == "__main__":
         with gr.Row():
             with gr.Column():
                 gr.Markdown(value="""
-                【AI向晚】在线语音合成（Bert-Vits2）\n
+                # 【AI向晚】在线语音对话版（Bert-Vits2 + gpt）\n
+                !(ava.png)\n
+                对话内容基于gpt生成，由于接口转发，生成可能需要2分钟！\n
                 作者：Xz乔希 https://space.bilibili.com/5859321\n
+                集成作者：碎语碎念 https://space.bilibili.com/4269384\n
                 声音归属：向晚大魔王 https://space.bilibili.com/672346917\n
                 Bert-VITS2项目：https://github.com/Stardust-minus/Bert-VITS2\n
+                GPT_API_free项目：https://github.com/chatanywhere/GPT_API_free\n
+                本项目中的apiKey可以从上面的项目免费获取（本项目默认提供了一个，如果没法用了去仓库申请替换就好啦）！\n
                 使用本模型请严格遵守法律法规！\n
                 发布二创作品请标注本项目作者及链接、作品使用Bert-VITS2 AI生成！\n                
                 """)
-                text = gr.TextArea(label="Text", placeholder="Input Text Here",
-                                      value="欸哇哇欸哇，这里是诶嗖的向晚，今天也要跟我一起轻松快乐的看直播。")
+                text = gr.TextArea(label="要对向晚说什么呢", placeholder="Input Text Here",
+                                   value="虚拟主播是什么？")
+                key = gr.Text(label="GPT Key", placeholder="请输入上面提示中获取的gpt key",
+                                  value="izlrijShDu7tp2rIgvYfibcC2J0Eh3uWfdm9ndrxN5nWrL96")
                 speaker = gr.Dropdown(choices=speakers, value=speakers[0], label='Speaker')
                 sdp_ratio = gr.Slider(minimum=0.1, maximum=1, value=0.2, step=0.01, label='SDP/DP混合比')
                 noise_scale = gr.Slider(minimum=0.1, maximum=1, value=0.5, step=0.01, label='感情调节')
                 noise_scale_w = gr.Slider(minimum=0.1, maximum=1, value=0.9, step=0.01, label='音素长度')
-                length_scale = gr.Slider(minimum=0.1, maximum=2, value=1, step=0.01, label='生成长度')
+                length_scale = gr.Slider(minimum=0.1, maximum=2, value=1.1, step=0.01, label='生成长度')
                 btn = gr.Button("点击生成", variant="primary")
             with gr.Column():
                 text_output = gr.Textbox(label="Message")
                 audio_output = gr.Audio(label="Output Audio")
-                gr.Markdown(value="""
-                【AI塔菲】https://huggingface.co/spaces/XzJosh/Taffy-Bert-VITS2\n
-                【AI东雪莲】https://huggingface.co/spaces/XzJosh/Azuma-Bert-VITS2\n
-                【AI奶绿】https://huggingface.co/spaces/XzJosh/LAPLACE-Bert-VITS2\n
-                【AI尼奈】https://huggingface.co/spaces/XzJosh/nine1-Bert-VITS2\n
-                【AI珈乐】https://huggingface.co/spaces/XzJosh/Carol-Bert-VITS2\n
-                【AI电棍】https://huggingface.co/spaces/XzJosh/otto-Bert-VITS2\n
-                【AI七海】https://huggingface.co/spaces/XzJosh/Nana7mi-Bert-VITS2\n
-                【AI阿梓】https://huggingface.co/spaces/XzJosh/Azusa-Bert-VITS2\n
-                【AI星瞳】https://huggingface.co/spaces/XzJosh/XingTong-Bert-VITS2\n
-                 """)
         btn.click(tts_fn,
-                inputs=[text, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale],
+                inputs=[text, key, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale],
                 outputs=[text_output, audio_output])
-    
+
 #    webbrowser.open("http://127.0.0.1:6006")
 #    app.launch(server_port=6006, show_error=True)
-        
+
     app.launch(show_error=True)