Video
Image
Audio
More
Discord

Turn an original lyric and production brief into a complete track with MiniMax Music 3, or switch models for older song, reference-audio, and instrumental workflows.

*
*
1300
1credit
Example Audio

MiniMax Music 3 — Neon Receipt

Generator examples

Real MiniMax Music 3 outputs · generated August 26, 2026

Listen before you write the brief

The fastest way to understand AI music generation is to hear what a model does with specific constraints. These three tracks use original lyrics, named arrangements, requested durations, and clean ending instructions. They are not stock audio or borrowed demos. A reliable AI music generation brief makes those choices visible before the first run. Use this AI music generation listening desk as a preflight before spending credits.

For a fully localized writing and review flow, use the Hungarian AI music generator or the Arabic AI music generator. Both keep the language-specific guidance beside the generator.

01

Neon Receipt

Meet me where the streetlights turn to gold

Synth-pop52 sec requested
02

Paper Maps

If the map runs out, we will draw the sea

Indie folk56 sec requested
03

Brightside ID

Brightside — make the next move clear

Pop jingle20 sec requested

Four useful routes, not one mystery dropdown

Choose the model by the control you need

An AI music generation model should be chosen by the control the assignment needs, not its position in the menu. In AI music generation, “best” changes with the assignment. Start with Music 3 for a new vocal song. Keep Music 1.5 for the older direct lyric workflow, use Music 01 when a reference track is central, and switch to MusicGen for instrumental scoring or continuation.

01Newest · default

MiniMax Music 3

Prompt + lyrics + duration

Choose it for
Complete vocal songs, longer structure, tightly directed arrangements
What you set
Duration, seed, guidance scale, and inference steps
02Song workflow

MiniMax Music 1.5

Prompt + structured lyrics

Choose it for
Full songs and concise sung jingles with verse/chorus form
What you set
Straightforward song brief with separate lyric structure
03Reference route

MiniMax Music 01

Style brief + reference audio

Choose it for
Original material shaped around a supplied musical reference
What you set
Use when the reference matters more than exact duration control
04Instrumental specialist

Meta MusicGen

Prompt, continuation, or melody guidance

Choose it for
Background scores, instrumental beds, continuations, and game ambience
What you set
Choose it when lyrics and a performed lead vocal are not the job

The selector above defaults to Music 3. Switch to an older model when the workflow, reference audio, or lyric structure you need is not available in Music 3.

Prompt and lyrics have different jobs

Write two documents, not one wall of text

In Music 3, Prompt directs the production; Lyrics directs the words and section order. Mixing both into one paragraph makes it harder to revise either layer. Keep the production brief concise enough to scan and the lyric structured enough to perform. For AI music generation with separate fields, this split is the simplest useful control.

Job complete song, underscore, hook, or jinglePulse tempo, groove, and energy curvePalette instruments, vocal role, and textureForm intro, verse, chorus, bridge, and outroMix intimate, wide, dry, dense, or voice-over safeEnding resolve, button, loop, fade, or edit point
music-3 / neon-receipt52 sec target
Prompt

Polished modern synth-pop, 112 BPM, warm expressive female lead vocal, glossy analog synth bass, tight electronic drums, muted guitar accents, intimate verse opening into a wide memorable chorus, crisp diction, emotional lift without over-singing, clean resolved ending.

Lyrics
[Verse]
Rain writes silver on the avenue
I kept one light on, waiting here for you
Every red sign fades into the blue
But the city keeps a neon receipt

[Pre-Chorus]
All the quiet hours add up tonight
One small spark can redraw the skyline

[Chorus]
Meet me where the streetlights turn to gold
We are more than all the plans we never told
Let the midnight open, let the old road go
Meet me where the streetlights turn to gold

[Outro]
Turn to gold, turn to gold

Duration is a structure decision

Build a song that lands on time

Music 3 accepts a duration request from one to 300 seconds, but the model may stop earlier when the song resolves. Do not solve timing with the number field alone. Give each section a job, put the hook where the edit needs it, and tell the model how to leave. In AI music generation, duration works best as a target plus a structural plan.

Cold open

Establish the palette before the vocal.

Verse

Keep diction exposed and the arrangement narrow.

Lift

Add motion without spending the full chorus early.

Chorus

Deliver the title line and the widest harmonic moment.

Outro

Resolve the last phrase and leave a usable edit point.

For a complete song

Spend the first seconds establishing the palette, then make the chorus easy to locate. A dense list of genres cannot replace an energy curve.

For a short asset

Put the vocal hook immediately, remove verse language, ask for one musical answer, and specify a dry resolved ending. The dedicated AI jingle maker turns that constraint into a focused workflow.

From idea to approved asset

AI music generation is a review loop, not a one-click release

An AI music generation output becomes usable only after it survives the target edit, device, rights review, and export path. Keep the winning brief with the audio so the next variant is an iteration rather than a restart.

  1. 01

    Define the job

    Decide whether the output is a complete song, a background bed, a repeatable loop, or a timed hook. If it sits under speech, leave room for the voice instead of asking for maximum musical detail.

  2. 02

    Try three contrasting takes

    Keep lyrics and structure fixed while changing one variable: tempo, vocal role, instrument palette, or intensity. Three clear directions teach more than ten loosely related prompts.

  3. 03

    Check it against the edit

    Judge a score inside the actual cut. Build motion with image to video, then check whether the hook, lift, and ending arrive where the visual does. A good standalone song can still be the wrong soundtrack.

  4. 04

    Save what worked

    Keep model, prompt, lyrics, settings, date, output, edit notes, and permissions together. If spoken copy must stay exact rather than sung, route it through text to speech and mix it separately.

Six passes before download

Review the track like an editor

Novelty is not quality. Listen once for each thing that can go wrong, then compare the result against the brief and where it will actually play. An AI music generation review should be as careful as any other edit in the project.

  1. 01

    Words

    Are the important lyric lines intelligible without reading them?

  2. 02

    Form

    Can you point to the verse, lift, chorus, bridge, and ending?

  3. 03

    Timing

    Does the useful moment arrive before the edit needs it?

  4. 04

    Mix

    Do vocal, bass, transients, and ambience survive phone speakers?

  5. 05

    Artifacts

    Listen for broken consonants, phantom syllables, clicks, or tails.

  6. 06

    Rights

    Record model, prompt, date, lyric ownership, and reference permission.

AI Music Generation FAQ

What is the newest model on this AI music generation page?

MiniMax Music 3 is the newest default model on this page. It accepts a production prompt, separate structured lyrics, and a requested duration from 1 to 300 seconds. The model can finish before the requested duration when it reaches a natural ending.

Can AI music generation create a complete song with vocals?

Yes. MiniMax Music 3 and MiniMax Music 1.5 are the song-first choices in the current selector. Give the model an original lyric, describe the vocal role and arrangement, and label sections such as Verse, Chorus, and Bridge on separate lines.

Which model should I use for instrumental background music?

Meta MusicGen is the specialist option for instrumental score, music continuation, and melody-guided soundtrack work. MiniMax Music 3 is the better starting point when sung lyrics and complete song structure are central to the brief.

How long can MiniMax Music 3 generate?

Its API accepts a requested duration from 1 to 300 seconds, so the upper request limit is five minutes. A request is a target rather than a guarantee: Music 3 may resolve the song and stop earlier.

Do I put the lyrics inside the music prompt?

Keep them separate when the model exposes both fields. Use Prompt for genre, tempo, instruments, vocal character, arrangement, mix, and ending. Use Lyrics for the exact original words and section tags.

What lyric tags work best for AI music generation?

Use simple structural tags such as [Verse], [Pre-Chorus], [Chorus], [Bridge], and [Outro], with each tag on its own line. Tags guide the arrangement; they do not guarantee an exact timestamp or identical repetition.

Can I ask the model to sound like a specific living artist?

Use musical attributes instead of asking for an imitation. Describe tempo, era, instrumentation, vocal register, rhythmic feel, production texture, and emotional arc. This produces a more useful brief and avoids leaning on someone else's identity.

Are the example tracks on this page real outputs?

Yes. The listening desk uses real MiniMax Music 3 outputs generated from the prompts and lyrics shown on the page. The browser previews are compressed for faster playback from the original stereo WAV results.

Can I use generated music commercially?

Usage depends on your plan, the active model's terms, the rights to every lyric or reference you provide, and the rules of the place where you publish. Keep the prompt, model, generation date, source material permissions, and final export with the project record.

Write the song as a production brief

Choose the model for the job, separate sound direction from lyrics, and judge the result on structure, intelligibility, mix, and ending — not genre alone.

Generate a track