How Dropkit turns a brief into a song

Dropkit takes the settings on the studio page, builds a style prompt and a lyric sheet from them, sends that brief to a music model, and then masters the result so it is ready for a DJ set. You never write the technical prompt yourself.

Step 1: your brief

The studio asks for a song name, an artist name, a vibe description of up to 2,000 characters, up to five genres with the first one leading, and lyrics. Lyrics can be yours, written for you, or skipped for an instrumental. You also set the groove: BPM from 60 to 200, a Camelot key, a length between 2:00 and 5:00, and an energy level.

The DJ mix edit section sets the arrangement: intro bars, outro bars, the bar of the first drop, break length, and switches for a drums-only intro, no vocals in the outro, a hard start on bar one, and a bass fade in the last eight bars.

Step 2: the prompt and lyric sheet

From those choices Dropkit writes one style prompt under the model's 1,000-character limit. It lists the genres, your vibe, any featured sounds, the BPM, the key by name, the energy, a sentence of vocal direction, and the DJ arrangement in bars. It closes with a request for a steady four-on-the-floor grid and a tight low end.

The lyric sheet is your lyrics with an [Intro] block marked instrumental in front and an [Outro] block marked instrumental behind. Any intro or outro sections you typed are removed so the sheet has exactly one of each.

Step 3: rendering

Dropkit submits the brief twice to the music model, so you get two takes, and submits the cover art request to an image model at the same time. A background worker checks every few seconds until each job finishes.

Step 4: mastering

When a take comes back, Dropkit splits it into vocal, drum, bass and other stems, mutes the vocal stem across the intro and outro windows, and sums the stems back together. Then it trims silence for a hard start, folds bass below 120 Hz to mono, normalises loudness to your LUFS target with a −1 dBTP ceiling, and writes a 320 kbps MP3 plus an optional 24-bit / 48 kHz WAV. Cue points are computed from your bar settings and the model's lyric timing.

Step 5: tags and delivery

Once the cover arrives, the MP3 is tagged with title, artist, genre, BPM, key in three notations, hot cues and the artwork. The track appears in your crate with a share page and a DJ pack download.