Lo-fi is the easiest genre to get a passable result in and the hardest to make interesting. The difference is arrangement: a loop that never changes reads as filler, so these blueprints all give the track somewhere to go — a break that drops out, an outro that fades.
A short instrumental with a deliberate arc rather than one repeated bar: the break drops out, the outro fades. Vocals are excluded explicitly, which is what stops stray humming from appearing.
Style
lo-fi hip-hop, slow tempo, around 78 BPM, Rhodes piano, bass guitar and live drums, lo-fi production, calm and nostalgic, soft vinyl crackle, muted rain in the background
Darker than the usual study loop without becoming dramatic. The palette stays at three instruments so the mix never crowds a voiceover sitting on top of it.
Style
lo-fi hip-hop, slow tempo, around 72 BPM, Rhodes piano, bass guitar and percussion, lo-fi production, calm and melancholic, brushed percussion, tape wobble on the keys
Three sections and a short duration, so it establishes a tone and stops. Ending on a defined beat rather than a fade gives you a clean point to start talking.
Style
lo-fi hip-hop, mid-tempo, around 88 BPM, Rhodes piano and acoustic guitar, lo-fi production, calm and hopeful, warm tape hiss, light finger-picked guitar
Pads instead of keys as the lead texture, which pushes the track further into the background. Useful when the picture is doing the talking and the music only has to set a temperature.
Style
lo-fi hip-hop, slow tempo, synth pads, Rhodes piano and percussion, spacious production, nostalgic and dreamy, blurred pad wash, brushes instead of a backbeat
How do I stop stray vocals appearing in an instrumental?
Add vocals to Exclude explicitly. Instrumental mode alone is a hint; the Exclude field is the instruction the model treats as a constraint.
Why does my lo-fi loop sound the same all the way through?
Because nothing in the prompt asks it to change. Give the arrangement a stripped-back section and a fade-out, and the model has a reason to vary the density.