To solve the problem of vocal and backing track dissonance, you can optimize it by following these steps:
- Using the multitrack output function: Add to the generate command
--separate_tracks
parameters, generating separate vocal and backing tracks for easy post-production balance adjustments - Precision Control Style Description: Ensure in the JSONL file that
descriptions
field contains a description of the matching rhythm (e.g.the bpm is 125
) and instrumental combinations - Check the lyrics segmentation: Strictly in accordance with
[verse]
/[chorus]
etc. structure to label lyrics, non-lyric passages (such as the[intro-short]
) should not contain textual content - Reference Audio Optimization: Upload a 10-second reference clip containing full vocals and backing vocals (chorus part is recommended), the model will harmonize the two better!
This answer comes from the articleSongGeneration: open-source AI model for generating high-quality music and lyricsThe