sovit dev

app.py

@@ -10,7 +10,7 @@ with app:
         with gr.TabItem("语音合成"):
             vits_inferencer = VitsInferencer("vits/configs/hoshimi_base.json")
             vits_inferencer.render()
-        with gr.TabItem("声线转换"):
+        with gr.TabItem("声线转换（开发中）"):
             sovits_inferencer = SovitsInferencer("sovits/configs/hoshimi_base.json")
             sovits_inferencer.render()
     app.launch()

sovits/sovits_inferencer.py

@@ -36,9 +36,13 @@ class SovitsInferencer:
         out_path = f"./out_temp.wav"
         soundfile.write(out_path, o_audio, self.svc.target_sample)
         mistake, var = self.svc.calc_error(audio_path, out_path, tran)
-        return f"分段误差参考：0.3优秀，0.5左右合理，少量0.8-1可以接受\n若偏差过大，请调整升降半音数；多次调整均过大、说明超出歌手音域\n半音偏差：{mistake}\n半音方差：{var}", (self.hps.data.sampling_rate, o_audio.numpy())
+        return f"半音偏差：{mistake}\n半音方差：{var}", (self.hps.data.sampling_rate, o_audio.numpy())
     
     def render(self):
+        gr.Markdown("""
+        未完成，效果有待提升\n
+        该模型适合**歌声**的声线转换，目前仅支持**45s以内**、**无伴奏**、**单声道**的**wav或mp3格式**文件
+        """)
         record_input = gr.Audio(source="microphone", label="录制你的声音", type="filepath", elem_id="audio_inputs")
         upload_input = gr.Audio(source="upload", label="上传音频（长度小于45秒）", type="filepath",
                                 elem_id="audio_inputs")