Feature Request: Support for backing vocals in generation (e.g. “(lyrics)” notation)

[saul246]

Hi 👋

First off, amazing work on SongGeneration, the demo sounds really impressive!

I was wondering if there are any plans to support backing vocals / background harmonies in generated songs.

For example, it would be really cool if the model could interpret something like:

Main line of lyrics (Backing vocals here)

It’d be great if the part in parentheses could be rendered as a separate backing layer instead of merging with the main vocal.

Is something like this on the roadmap?

Thanks again for sharing this project with the community 🙌

[tencent-ailab]

This is a good idea, we will take it into consideration in the subsequent plan. However, due to the difficulty in obtaining data, we have not yet started.