Lofi is a forgiving genre to start with because its defining features — slow tempo, jazzy harmony, imperfect texture — are easy to name explicitly in a prompt, and the genre tolerates the slightly loose, hazy quality AI generation sometimes has anyway.
Step 1: write the core prompt
Name the genre, the tempo range and the mood directly: 'lofi hip-hop, 75 BPM, warm and nostalgic, jazzy electric piano chords, dusty boom-bap drum loop, no vocals.' The 'no vocals' instruction matters — leaving it out sometimes produces mumbled or wordless vocal textures you may not want for a study-beat use case.
Step 2: generate variations and pick the loop-friendliest one
Generate three or four variations from the same prompt rather than accepting the first result. Listen for which one has the most consistent groove throughout — lofi's appeal is in its steady, repeatable feel, so a version with a distracting one-off fill or a sudden dynamic swing is harder to use as a loop, even if it sounds fine as a one-off listen.
Step 3: ask for the texture in the prompt
The genre's signature 'dusty' character often comes from ambient layers under the music, not the music itself. Name it in the style prompt — for example 'vinyl crackle, soft rain ambience' — and the texture is generated with the beat. If it comes out too loud or not at all, generate a variation rather than rewriting the whole prompt.
Step 4: trim to a clean loop
If you need a beat that loops seamlessly (for a background stream, a game menu, or a long study playlist), use the cut tool to trim the track at a bar boundary rather than an arbitrary point, and check the loop by listening across the seam a few times before finalizing.