MusicGen: SOTA Controllable Text-to-Music Model

Any music people able to give comments on how much prompt engineering is needed?