ILG2021
ILG2021
still needs the feature.
I have implement it by ffmpeg: ``` output_audio_file = "output_audio.mp3" ffmpeg_command = ["ffmpeg", "-y", "-i", temp_audio_file, "-filter:a", "atempo=1.3", output_audio_file] subprocess.run(ffmpeg_command, stdout=subprocess.PIPE, stderr=subprocess.PIPE) ```
指定language=‘zh'，然后我用prompt是'你好，晚上好，很高兴认识你。'，能够出来简体。
Yes, WhisperHallu is a very good solution for the silence sound which cause youtube ads. But it seem can not solve the repeat sentence problem. I have a Chinese fine...
> @ILG2021 Thanks ! With a lot of not significant differences (spaces), it's hard to figure out what were exactly the pertinent parts of your modifications. I think I will...
Can you use pycharm? so it will recognize changes quickly.
> 这个不同性别测试的有点魔幻，请问是在测试前对输入的歌曲进行变调处理还是测试后进行变调处理会比较好呢。测试前进行变调处理，我用的软件，这样比较麻烦。但是效果还可以。测试后进行变调处理感觉和原声区别比较大。 请问下变调使用什么软件？
使用--shift 变调，效果还可以
你手动创建logs\sovits5.0 看看 我用pycharm+venv，训练没有问题。
someone has implemented the [facodec](https://github.com/Plachtaa/FAcodec), I expected that we can implement natural speech 3. @lucidrains