How to build an AI music prompt with a selector: brief, sound and arrangement
A selector is useful when each choice has a purpose. Filling every field makes a prompt longer; it does not automatically make the music clearer. This exercise produces one saved blueprint, one compiled prompt and one specific question for your first listen.
1. Write the job before choosing the sound
Open Studio and create a project. Choose vocal or instrumental mode. In the creative brief, select a use case and an energy arc, then write two sentences about the scene and the constraint that matters most.
For example: “A quiet opening for a walking video. The music should leave the first spoken sentence clear.” Keep the duration requirement in your project notes as well. Brief notes guide your decisions and optional AI writing; the local compiler does not turn them into sung lyrics or automatically insert them into Style.
For a vocal song, put the actual words in Lyrics. A line such as “the chorus should feel larger” belongs in a musical direction, not among words intended to be sung.
2. Select only the decisions you can explain
Choose one primary genre. Add a secondary genre only if you can describe the contrast you want. Select up to two moods and three instruments; a small set makes it easier to notice when the generator adds something unwanted.
The key, time signature and groove selectors add English musical directions to Style in either interface language. BPM, key and meter are requests, not guarantees that the resulting audio will be locked to them. Leave a value blank when you do not need it.
An original instrumental starting point is:
| Decision | Selection | What to check |
|---|---|---|
| Genre | Ambient | Does it stay out of the narrator’s way? |
| Instruments | Piano, synth pads | Is piano still the foreground? |
| Mood | Calm | Does a sudden accent disturb the scene? |
| Groove | Free rhythm | Does a strong beat appear anyway? |
| Production | Minimal | Are extra layers making the mix busy? |
These are exercise choices, not a measured preset. The search box accepts English and Chinese labels: try piano or 钢琴 and compare the available choices.
3. Give sections a role
For a vocal song, choose a verse–chorus or short-hook starting structure. For an instrumental, try a gradual build or a quiet ambient arc. Applying a structure creates a new version. Matching sections keep their text; unmatched written passages are retained before End, and newly added slots remain empty.
Read every retained passage before generating. A template can move a section while preserving its words, so the narrative may now need a transition. Use local directions such as a restrained verse followed by a fuller chorus instead of asking every section to be energetic.
4. Inspect the compiled result
The live preview shows Style, Exclude and the arrangement as you change choices. Look for contradictions before exporting: requesting a drum part while excluding drums is an error, not an interesting blend.
Choose the model you will actually use. New projects use the documented v6-mini label; v6 slider mappings are not yet live-interface verified, so configure those settings in Suno. The model guide explains the distinction.
Compile when the preview is coherent, then follow the copy guide. Save the blueprint before requesting audio.
Finish the exercise
Listen with the actual video or vocal context. Write one timestamp and one observation, then choose one change for the next version. Your deliverable is a pair of blueprints with a clear difference, not a folder of unrelated generations. Continue with the listening journal or explore the full learning path.