zmime.com
Category · Audio

Audio prompts

Music and voice models respond to arrangement notes, tempo and mix language — not to genre names alone.

  • Music
  • Voiceover
  • Sound design
  • Podcast
  • Sonic branding
01 — Overview

Why these prompts look the way they do

A genre label leaves almost everything undecided. Tempo, key, instrumentation, arrangement over time, and mix character are the parameters that decide whether a track fits your video. The same is true for voice: "friendly" is not direction, but "warm, 145 words per minute, breath before each key phrase" is.

Audio prompts also need a purpose. A 15-second ad bed, a loopable background for a 10-minute tutorial and a title sting are three different briefs even if the genre is identical.

02 — Anatomy

What an audio prompt must specify

  • Purpose + length Where it plays and for how long. Loopable or one-shot.
  • Tempo + key BPM and key or mode. Removes most of the randomness.
  • Instrumentation Named instruments and their roles, not just a genre.
  • Arrangement What happens at 0s, mid-point and end. Builds need a timeline.
  • Mood + reference Emotion plus a descriptive reference rather than a copyrighted artist.
  • Mix character Dry or reverberant, wide or narrow, low-end weight, loudness target.
  • Exclusions No vocals, no drums, no risers — whatever must not appear.
03 — The prompts

5 copy-ready audio prompts

Replace everything in [brackets]. Keep the constraint lines — they do more work than the descriptive ones.

Nothing matches that word. Clear the field to see every prompt.

Prompt 01

Background music bed for video

Under-narration music that does not fight the voice.

  • music
  • video
Template · plain text
Instrumental background bed for a [duration] [type of video], designed to sit under narration.

TEMPO & KEY: [90] BPM, [C minor], steady, no tempo changes.
INSTRUMENTATION: soft electric piano as the main voice, warm sub bass, light brushed percussion, one sustained pad. No lead melody that competes with speech.
ARRANGEMENT: 0-4s soft entry with pad only; 4s bass and percussion join; hold a consistent groove throughout; last 4s decay to pad and a single piano note.
MOOD: focused, optimistic, unhurried. Neutral enough to loop.
MIX: mid-forward instruments kept quiet, no sharp transients, gentle high-shelf roll-off, stereo but centred low end, mastered around -16 LUFS.
EXCLUDE: vocals, risers, cymbal crashes, sound effects, key changes.

Deliver a seamless loop with matching in and out points.
Prompt 02

Voiceover direction script

Feeding a TTS model or briefing a voice actor.

  • voiceover
  • tts
Template · plain text
Read the script below as a [age range] [accent] narrator.

VOICE: [warm / crisp / gravelly], medium-low pitch, relaxed.
PACE: [145] words per minute. Slow by 15% on any sentence containing a number.
DELIVERY NOTES
- Read as if explaining to one person across a table, not to a crowd.
- Small pause (0.4s) at each paragraph break, marked below with [pause].
- Emphasise the words wrapped in *asterisks*.
- Upward inflection only on genuine questions.
- No smiling tone on the caution sentence.

SCRIPT:
"""
[paste script with *emphasis* and [pause] markers]
"""

Output: clean read, no added words, no music, consistent room tone.
Prompt 03

Sound effect design brief

UI sounds, foley and transitions.

  • sfx
  • ui
Template · plain text
Design a single sound effect: [what the sound represents].

LENGTH: [0.4] seconds, one-shot, silence at the start and end.
CHARACTER: [material] impact with [texture] tail, [bright/dull], [organic/synthetic].
LAYERS: 1) transient click for definition, 2) body tone around [frequency], 3) short tail with [reverb type], under 200ms.
DYNAMICS: fast attack, controlled decay, no clipping, peaks at -6 dBFS.
CONTEXT: plays [when], must be audible on phone speakers and not fatiguing when heard 50 times an hour.

EXCLUDE: musical pitch that clashes with our [key] brand theme, harsh 2-4kHz, long reverb.

Provide 3 variations of increasing brightness.
Prompt 04

Podcast episode production plan

Turning a raw recording into a structured episode.

  • podcast
  • editing
Template · plain text
You are a podcast producer. Using the transcript below, plan episode [n] of [show], a [duration] show for [audience].

DELIVER
1. Episode title options (5, under 60 characters, no colons).
2. Cold open: pick the strongest 20-second quote, with timestamps.
3. Segment map: timestamp ranges, segment name, one-line purpose.
4. Cuts: list tangents, repetitions and dead air worth removing, with timestamps and reason.
5. Chapter markers with timestamps.
6. Show notes: 120-word summary, 5 bullet takeaways, list of every resource actually mentioned.
7. Three social clips: timestamps, why each works, suggested caption under 200 characters.

RULES
- Quote the transcript exactly for anything in quotation marks.
- Never invent a resource, name or statistic that is not in the transcript.

TRANSCRIPT:
"""
[paste transcript with timestamps]
"""
Prompt 05

Sonic logo / brand sting

A short signature sound for intros and app launches.

  • branding
  • sting
Template · plain text
Create a sonic logo for [brand], a [industry] brand described as [3 adjectives].

LENGTH: 1.8 seconds maximum.
MOTIF: [3-4] notes, rising, in [key], memorable when hummed.
SOUND PALETTE: [e.g. glass marimba plus a soft analog pad], one percussive accent, no drums.
SHAPE: quick attack, brief sustain on the final note, tail resolving into silence by 1.8s.
MIX: clean and dry with a hint of short plate reverb, works on laptop speakers and headphones, no sub-40Hz content.
VARIANTS: deliver 3 versions — full, a 0.8s short form, and a single-note notification variant using the same timbre.

EXCLUDE: vocals, whooshes, orchestral hits, anything resembling a well-known existing sting.
04 — Craft notes

What separates a good prompt from a wasted run

Do this 05

  • State BPM, key and length in every music prompt.
  • Describe the arrangement as a timeline, not a mood.
  • Name instruments and give each one a role.
  • Set a loudness target when the audio sits under speech.
  • Ask for variants so you can A/B in context.

Not this 05

  • Naming a specific artist or track to imitate.
  • Genre-only prompts like "make it lo-fi".
  • Melodic leads under narration.
  • Long reverb tails on UI sounds.
  • Judging a bed in isolation instead of against the video.
05 — Questions

Asked often, answered plainly

Can I reference a real song for the vibe?

Describe the musical characteristics instead — tempo, instrumentation, groove, mix. Naming an artist invites imitation you may not be able to license, and most tools restrict it anyway.

How do I stop generated music from overpowering narration?

Ask for a mid-scooped, low-transient bed with no lead melody, then set a loudness target such as -16 LUFS. Always audition it under the actual voice track rather than on its own.

What makes text-to-speech sound robotic?

Punctuation and pacing. Write for the ear: shorter sentences, commas where a person would breathe, explicit pause markers, and numbers spelled the way you want them read.

06 — Sources

Read the primary documentation

Vendors change flags, limits and defaults often. Where this site disagrees with official docs, the docs win.