Answers for your next song idea.

Which answer applies to your project?

Use the FAQ as a decision map, not a wall of definitions.

If you are still choosing a direction, start with the difference between a song and a cue. A complete song benefits from a story, lyrics or vocal intent, and section movement. A background cue benefits from a clear edit role, controlled density, and enough space for speech or other sound. The right prompt becomes shorter once that job is clear.

If you are deciding whether to create, check the current credit and storage details first. One complete song uses 4 credits in the current Studio workflow, while the Free plan is presented as one complete song daily. Paid plans add more monthly capacity and eligible commercial usage rights; the plan attached to the generation matters when a track becomes client, advertising, monetized, or other public work.

If you already have a first result, do not ask whether it is universally good. Ask whether the vocal, rhythm, texture, duration, and output format serve the intended listener. Keep the strongest choice, change one weak detail, and compare the next version with the same project goal. That review loop is usually more useful than adding more adjectives to the original prompt.

A decision guide

Use the FAQ to choose the next action, not just the next answer.

If you are still choosing a direction, start with the difference between a complete song and a supporting cue. A song needs a point of view, lyric or vocal intent, and movement between sections. A cue needs an edit role, controlled density, and a clear ending or loop. Once the job is clear, the prompt becomes shorter because you no longer ask one track to solve several different problems.

If you are deciding whether to create, check the current credit and storage details before you commit to a workflow. One complete song uses 4 credits in the current Studio flow. The Free plan is presented as one complete song daily, while paid plans add monthly capacity and eligible commercial usage rights. The plan attached to the generation matters when the result becomes client work, an advertisement, a monetized video, or another public asset.

If you already have a first result, ask whether the vocal, rhythm, texture, duration, and output format serve the intended listener. Keep the strongest choice and change one weak detail. This review loop is more informative than adding adjectives without listening. Use the Music GPT guide for the prompt workflow, the Studio for a working generation, and the pricing page plus usage-rights guide before publishing.

The short answer to “what can MusicGPT make?” is: a complete vocal song, an instrumental direction, a lyric-led idea, or a background cue described in ordinary language. The useful choice is the job of the output. A song needs a point of view and section movement; a cue needs timing, density, and space for another message.

The short answer to “what should I put in a prompt?” is: name the listener, scene, emotional movement, two or three sound anchors, vocal or instrumental mode, and one constraint. Add a language, duration, loop, or output-format detail only when it changes the result you need to review.

The short answer to “is the first result ready?” is: listen in context. Compare it with the picture, voice, lyric, or audience it is meant to serve. Keep the strongest musical decision, make one focused revision, and check the current plan and supplied-material rights before a public release.

The short answer to “song or background music?” is to choose the role before the style. A song carries a vocal story and a memorable section; background music supports speech, picture, or a product moment. If the same prompt tries to do both, the first result may be harder to judge because its arrangement has no single priority.

The short answer to “how do I compare versions?” is to keep the destination constant. Listen to the same edit, lyric, or voiceover with each result, note the opening, density, emotional change, and ending, then revise the one part that does not serve the listener. A consistent test is more informative than a larger pile of unrelated generations.