问HN：这些AI翻唱歌曲是如何制作的？
最近在YouTube和Instagram上出现了一波关于人工智能的音乐翻唱。例如： https://youtube.com/@fakemusicbr?si=G-RzKcm2iuXX56kI
其中一些歌曲真的很不错，远超我对音频人工智能的预期。有人知道使用了哪些模型和工具吗？我想一旦你有了音轨，可能就可以进行风格/流派转换，并使用参考音频文件，就像图像和视频模型一样。
查看原文
There's been a recent flurry of AI covers on Youtube and Instagram. For example: https://youtube.com/@fakemusicbr?si=G-RzKcm2iuXX56kI<p>Some of these songs are _really_ good. And pretty far ahead of where I thought audio AI was. Does anyone know what models and tools are used? I imagine once you have stems you can maybe style / genre transfer and use a reference audio file perhaps? Like image and video models.