AI covers: RVC (MIT, still updated) leads as rivals archiving

The article treats AI song covers not as a single product but as a small pipeline, and that framing explains why the details matter. The recognisable result depends first on how cleanly the original vocals can be pulled apart from the backing track, and tools that handle this well are grouped under names like MDX-Net, Demucs and RoFormer. What follows—voice conversion—is where the character of the finished cover is decided, and the piece points to RVC's retrieval-based approach as the reason the original singer's voice tends not to bleed through.

The maintenance status of these tools tells its own story. RVC is MIT-licensed and still updated, while Seed-VC and so-vits-svc have been archived. For readers who want to try the workflow without building it themselves, the article suggests browser services, and then sets out the questions to ask of any of them: where the voices come from, what happens to an uploaded song, and what the free tier actually allows.

Rights come last but frame everything. The song and the voice carry separate permissions, and publishing or monetizing a cover can require both, which suggests the practical ceiling on this hobby may be legal rather than technical.