How long can MusicGen generate in one pass?

MusicGen reliably produces up to about 30 seconds of audio per generation. The model can technically be pushed longer, but coherence degrades, so for longer pieces you chain generations using continuation mode.

What is continuation mode?

Continuation feeds the model an existing audio segment and asks it to keep going in the same style. It is how you build coherent tracks longer than 30 seconds — generate a base clip, then continue from its final seconds repeatedly.

What is melody conditioning?

The melody-conditioned MusicGen variant accepts a reference melody (a hummed or played line) alongside the text prompt. The model then generates accompaniment in your described style while following that melodic contour, which is useful for arranging a tune you already have.

Does MusicGen understand BPM and key?

It responds well to tempo cues like "120 BPM" or "slow" and to mood-implied keys ("melancholic", "uplifting"), but it does not take an explicit musical-key parameter. Describe the feel rather than naming a specific key signature.

MusicGen Prompt Guide

MusicGen prompt guide

MusicGen is Meta’s open text-to-music model (part of AudioCraft). It is text-conditioned: you describe the music in words and it generates audio. It is fast and runs locally, but it has two quirks worth planning around — a practical 30-second per-generation limit and several modes (plain generation, continuation, and melody conditioning) that each call for a slightly different prompt approach.

How it works

In generation mode, MusicGen reads a short descriptive phrase — genre, instrumentation, tempo, mood — and produces a clip up to roughly 30 seconds. For longer tracks you switch to continuation: you pass in an existing audio segment and the model extends it in the same style, which keeps a multi-minute piece coherent. The melody-conditioned variant additionally takes a reference melody and arranges your described style around that contour. The builder here writes the text-conditioning string and reminds you which mode fits your goal.

Tips for better MusicGen output

Be concrete about instruments. “warm Rhodes piano, soft brushed drums, upright bass” gives a far more controlled result than “jazzy”.
State tempo, not key. “90 BPM, relaxed” works; an explicit key signature does not, so describe mood instead.
Chain with continuation for length. Don’t push a single call past 30 seconds — generate a base and continue from its tail to stay coherent.
Use melody conditioning to arrange. If you already have a tune, the melody-conditioned model will dress it in your described genre.