Module 6 of 6 · Intermediate · 10 min read

Music, audio and voice

Generate music, sound effects and voice-overs, clean up recordings, and stay on the right side of consent and licensing.

Key takeaways

1Describe music by genre, tempo, mood, instruments, structure and vocals.
2Use AI cleanup and transcription to speed up podcast and video audio.
3Only clone voices with clear, documented consent.
4Check commercial-use terms before using generated music or voices in published work.

01Where AI helps

Audio tools have improved quickly; quality is often good enough for social, games and prototypes, with professionals still preferred for flagship work.

Generating songs, background music and stems from text.
Sound effects and ambience for videos, games and apps.
Text-to-speech voice-overs in many voices and languages.
Cleaning recordings: noise removal, levelling, separating voice from music.
Transcription, captions, translation and dubbing.
Lyrics, song structure ideas and production feedback.

02The workflow

Short, specific prompts with a clear structure usually beat long, poetic ones.

Define the use: length, mood, where it will play, and whether it needs vocals.
Write a prompt or lyrics; generate several versions.
Pick the best, extend or regenerate sections, and export stems if available.
Mix in an audio editor or DAW: levels, EQ, fades, timing to picture.
For voice, write the script for the ear, generate, then fix pronunciation.
Check licences and disclosure requirements before publishing.

03Tools to reach for

Licensing terms differ widely between tools and plans; they matter as much as quality.

Text-to-music generators, some with lyrics, vocals and stem export.
Text-to-speech and voice design tools, with pronunciation controls.
Sound-effect generators and AI audio cleanup and stem-separation tools.
Speech-to-text models for transcripts and captions.

04Prompt: music

Give the musical facts a producer would need.

Example

Instrumental background track for a product demo video, 60 seconds. Upbeat but calm indie electronic, around 110 BPM, major key. Instruments: soft plucked synth, warm bass, light electronic drums, airy pads. Structure: 8-bar gentle intro, build at 0:20, full groove from 0:30, clean ending at 0:58. No vocals. Leave space in the mid-range for a voice-over.

05Prompt: a voice-over script

Write for listening: short sentences, natural rhythm, pronunciation guides.

Example

Rewrite this text as a 45-second voice-over script for {{audience}}. Spoken style: short sentences, no parentheses or lists, numbers written as they should be read aloud. Mark pauses with '...' and emphasis with CAPS on single words. Add a pronunciation guide for any brand names or technical terms. <text>{{text}}</text>

06Prompt: sound design brief

Layer sounds by describing them separately.

Example

Create three short sound effects for a mobile game's coin pickup, each under 1 second: 1) bright, sparkly chime with a quick upward pitch sweep; 2) a softer wooden 'tok' with a tiny shimmer; 3) a chunky retro 8-bit blip. Dry, no reverb, no music, normalised so they sit at the same loudness.

07Pitfalls

Consent and licensing are not optional extras in audio. Keep records of both.

Voice cloning without consent: unethical, often against tool terms, and illegal in some places.
Sound-alikes: prompts naming famous artists or songs can create infringement risk. Describe the style instead.
Licensing: check whether your plan allows commercial use and monetisation, and what credit is required.
Robotic delivery: rewrite scripts for the ear and adjust pacing and pronunciation.
Loudness: match levels between generated clips and your other audio before publishing.

Watch out

Never use a cloned voice to impersonate a real person or to make it appear someone said something they didn't. It can cause real harm and may be illegal.

Finished reading?

Mark it complete to track your progress, then test yourself with the Build with AI: Software & Media quiz.

Take the Build with AI: Software & Media quiz →