How to write a music brief an audio tool will actually follow

You have lyrics. You paste them into an audio tool with a line like "happy kids song, ukulele" and it hands back something at 140 beats per minute, sung by an adult woman with a lot of reverb, in a key no child can reach. You try again. It hands back a different wrong song.

The tool did what you asked. "Happy kids song" describes several thousand recordings, and it picked one. A music brief is the short document that removes the guessing.

The six things a brief has to state

1. 01

   Tempo, as a number

   A rhyme a small child sings along to sits between 90 and 110 beats per minute. Under 85 it drags. Over 120 the words stop being sayable. Write the number, not the word "upbeat".

2. 02

   Key and range

   Children's voices live roughly between middle C and the A above it. Ask for C, D or F major and a melody that stays inside one octave. Left unsaid, tools default to adult ranges.

3. 03

   Who is singing

   State it plainly: a small children's choir, or a single child voice, or a warm adult voice leading with children answering. This one decision changes the whole feel and it is the one most briefs omit.

4. 04

   Instruments, and a ceiling

   Name three or four and say no more than that. Ukulele, glockenspiel, soft hand percussion, upright bass. A ceiling matters more than the list, because a dense mix buries the words.

5. 05

   Structure, spelled out

   Chorus first, then verse, chorus, verse, chorus, and a short tag. Write it as a line the tool can read. Most tools respect an explicit structure and invent one otherwise.

6. 06

   What to avoid

   No key changes, no rap sections, no heavy reverb, no vocal runs, no drop. Negatives do real work here, because the training data is full of adult pop conventions.

What that looks like written down

Every kit ships with a music brief in this shape, because the shape is the part that transfers between tools. Paste it above the lyrics.

Children's nursery rhyme, 100 bpm, D major.
Small children's choir, unison, melody inside one octave.
Ukulele, glockenspiel, soft shaker, upright bass. Nothing else.
Structure: chorus, verse 1, chorus, verse 2, chorus, 4-bar tag.
Warm, close, dry. Words in front of the mix at all times.
Avoid: key changes, vocal runs, reverb washes, rap, drop, sound effects.

Six lines. It is not a longer prompt than the one that failed, it is a more decided one. Every line closes a door the tool would otherwise walk through.

Fix the tempo before you fix anything else

If a returned song feels wrong and you cannot say why, it is almost always the tempo. Say the chorus out loud at a speed a four-year-old could copy, count the seconds for four lines, and you have your number.

This also decides your video. A 100 bpm song with four-line verses gives you a cut roughly every seven seconds, which is the shot length a small child can actually read.

Keep the take you can build on

Generate three or four takes from the same brief and judge them on one question only: are the words clear? Not which is prettiest. A pretty mix with a mumbled chorus is unusable, and no amount of editing later puts the words back.

Then keep the brief. The next rhyme for the same channel should use the same tempo band, the same voice and the same four instruments, because that consistency is what makes a series sound like a series.

One caution about wording. Ask for a named artist or a specific existing song and most tools will refuse, and the ones that do not will hand you something you should not publish. Describe the sound instead.