Cover
If you are using a beta version of MelonSound then you will have access to the Cover toggle. Cover allows you to select a music file on your computer and have MelonSound create a cover based on it.
When creating covers all other parameters work just like before, but the new song will be based on the provided cover audio.
Using covers you can convert an existing song into a new style. For example, a rock song can be converted into an unplugged aucoustic live session, or a a country song can be converted into jazz.
Structure such as sections, energy arc and melody shape are preserved while instruments, vocal character and timbre are changed.
Strength
The strength slider allows you to control how strictly the provided audio file should be followed. The higher the value the less room you give to MelonSound to make changes.
Below you will find a rough guide of how to think about the cover strength
| Action | Range |
|---|---|
| Very loose Style transfer | 0.1 - 0.3 |
| Dramatic Change (country to metal) | 0.3 - 0.5 |
| Moderate Change (pop to jazz) | 0.5 - 0.7 |
| Subtle Transformation (rock to indie rock) | 0.7 - 0.9 |
| Only Minor Slteration (no genre change) | 0.9 - 1.0 |
Lyrics
If you do not provide any lyrics when generating a cover then the output can be random. Sometimes you may get the same lyrics, and sometimes you may end up with something completely different. It is recommended to always provide lyrics, even if they are the original ones. This clearly indicates to the underlying AI model what you want it to do.
You can provide entirely new lyrics, and the underlying AI model will try to fit those into the structural timing of the provided music track. This means that you can create parodies, translations or change the theme of a song while keeping the melody intact. As a general rule of thumb you want to keep the structure of your lyrics in line with the structure of the provided song. If you struggle with singing the provided lyrics over the provided audio track then the AI will too.