Classic Shanghai jazz / soul
Female vocal·Mandarin·2:46
NewMiniMax Music 3 is live on Mai Music — 15 credits free when you sign up. See plans
Mai Music AI Music Generator
Describe the song you want and write the words. MiniMax Music 3 composes, arranges, performs and produces it in a single pass — up to five minutes of finished music with vocals, not a thirty-second fragment to stitch together.
The composer opens on MiniMax Music 3 — switch any time
Genre · tempo · key · mood · voice · arrangement
1:00
0:155:00 max
Model: MiniMax Music 3
The composer above is the one inside Mai Music, running on this page. A short song usually comes back in about a minute; a five-minute one takes longer. The exact credit cost is shown on the Create button before you spend anything.
Made with MiniMax Music 3
The demo library
Sixteen songs, and the briefs that made them
These are the 16 demos MiniMax released with Music 3.0, grouped the way MiniMax grouped them — each one published with the brief and the lyric sheet that produced it. Play any of them, then load its brief into the composer above and change a line.
Long-form4Prompt adherence3Arrangement3Audio quality3Vocals3
The four songs MiniMax published alongside the architecture
Futuristic melodic EDM / progressive house
Lead vocal·2:45
Progressive house / EDM
Male tenor·2:49
Bright power pop / pop rock
Female mezzo-soprano·2:51
Generated by MiniMax for the Music 3.0 release of 13 August 2026 and licensed for use on Mai Music. Not generated on this site — every file carries MiniMax's signed AIGC label naming them as the producer.
The two inputs
A brief and a set of lyrics. That is the whole interface.
MiniMax Music 3 reads two fields. The lyrics decide what is sung and where the sections fall; the brief decides everything else — genre, tempo, key, the voice, the instruments and how the arrangement grows. Written the way MiniMax documents it, a brief has three parts, and the difference between a thin result and a good one is almost entirely in here.
The brief
Global: synth pop. BPM: 104. Key: A minor. Nocturnal and intimate, opening restrained and widening at each chorus. Vocal: soft female lead, close and breathy, doubled an octave up on the chorus with light stacked harmonies. Arrangement: analogue pad and muted electric guitar under the verse; gated drums and a round synth bass enter at the pre-chorus; the last chorus adds a counter-melody on a bright lead. Wide reverb, tape saturation on the drum bus.
The lyrics, with section tags
[Intro] [Verse] The window keeps the city out A hum of lights I never doubt You said the quiet suits me now [Pre-Chorus] And every street is holding still [Chorus] Neon quiet, neon quiet Everything I meant to say Turning over in the light [Verse] A cab goes past and takes the sound I count the floors on the way down [Bridge] If it stays this loud inside my head I'll take the long way home instead [Chorus] Neon quiet, neon quiet Everything I meant to say Turning over in the light [Outro]
It fills the form at the top of the page. Edit anything before you generate.
Naming the three parts of the brief is most of the skill. A one-line prompt works; a structured one works better.
One pass, not a stitch
Five minutes with the same voice at the end as at the start
Most AI music generators hand you thirty-second fragments and leave the assembly to you, which is why the vocal changes character halfway through. MiniMax Music 3 writes the whole song at once, so a theme introduced in the first verse can come back in the last chorus and the singer is recognisably the same person throughout.
0:002:305:00
[Intro][Verse][Pre-Chorus][Chorus][Verse][Bridge][Instrumental][Chorus][Outro]
The bar above is a conventional pop structure drawn to scale — an illustration, not a contract. Section tags and a brief give MiniMax Music 3 direction rather than a score, so the tempo, the key and the exact bar a chorus lands on can come back different from what you asked for. Generate twice and keep the better take; that is what the seed control is for.
Every tag the lyrics field accepts
Nine tags mark where each part of the song begins, and the model builds the arrangement around them. That is what makes a five-minute generation hold together rather than wander.
[Intro][Verse][Pre-Chorus][Chorus][Post-Chorus][Bridge][Instrumental][Solo][Outro]5:00
Longest song in a single generation
9
Section tags you can arrange
8B
Parameter model holding the song's long-range structure
5,000
Tokens of brief the model will read
What to ask for
Four things this AI music generator does today
Not a roadmap. Every one of these is in the form at the top of this page, on MiniMax Music 3, the model it opens on.
A complete song
Text to music with vocals: words, arrangement and production in one generation. Bring your own lyrics, or have them written from your brief and edit what comes back.
Try itInstrumental passages
There is no vocals-off switch. Write [Instrumental] or [Solo] where the voice should step out, and the arrangement fills the space.
Try itThe length you asked for
Anything from a few seconds to a five-minute track, set on a slider. Credits scale with duration, so a sketch costs a fraction of a finished song.
Try itThe same song twice
Set a seed and the run repeats, so you can change one line of the brief and hear only that change. Inference steps and guidance are there when you want them.
Try itTwo music models
MiniMax Music 3 by default, Suno when you need it
Default
MiniMax Music 3
Text to song: complete tracks from a brief and lyrics, up to five minutes long, with the structure you wrote.
- One pass to a finished five-minute arrangement
- Section tags put the verses and choruses where you want them
- Open weights, published under MiniMax's community licence
Suno v5.5
The wider toolkit: reference audio, saved voices, stems, covers and extensions.
- Two takes per generation, scored separately
- Split a finished song into its parts, or extend it
- Save a voice and reuse it across songs
What Mai Music is
A MiniMax Music 3 AI music generator that runs in the browser
Mai Music is a MiniMax Music 3 AI music generator you can use without installing anything — text to music in a browser tab. The model itself is a large one — an eight-billion parameter language model handling long-range musical structure, a smaller one handling frame-level detail, and a flow-matching synthesiser turning the result into audio — and running it yourself means a CUDA card with tens of gigabytes of memory. Here it is a form in a web page: you type a brief, you type or generate lyrics, you set a length, and a finished song comes back to your library a minute or so later.
Generation is billed in credits and the cost scales with the length you ask for, so a fifteen-second sketch is cheap and a five-minute track is not. Signing up puts 15 credits in your account — enough for a full sixty-second song, or three short sketches — and everything you generate stays in your library with its brief and lyrics attached. Every track can be downloaded as a file.
Write the brief the way MiniMax Music 3 reads it — a one-line prompt works and a structured one works better. Naming global metadata, vocal details and arrangement is most of the skill.
Section tags are how you arrange a song. Nine tags mark where each part begins and the model builds the arrangement around them; leave a section tagged [Instrumental] and the singer sits out while the band plays.
Credits, downloads and what you own. Costs scale with length, everything stays in your library, and MiniMax publishes Music 3 under its own community licence — read it before you build a business on the output.
Two music models. The composer opens on MiniMax Music 3; switch to Suno v5.5 when you need reference audio, saved voices, stems or covers.
AI music guide
How AI music generators turn text into songs
What is an AI music generator?
An AI music generator is a tool that composes, arranges and produces a complete song from text. You describe the music you want — genre, mood, tempo, key, the voice, the instruments — and the AI music model handles the rest. The result is a finished recording with vocals, not a loop or a sample pack.
The best AI music generators work from two inputs: a brief and lyrics. The brief describes the sound; the lyrics decide what is sung and where the sections fall. That text to music pipeline is what makes it possible to write a song in minutes instead of days.
On Mai Music, the AI music generator runs in your browser. Type a brief, write or generate lyrics, pick a length up to five minutes, and a complete song with vocals, arrangement and mix comes back to your library. No music theory required.
How to make a song with AI: text to music in four steps
Making a song with AI music tools is a four-step process. Everything below is what the Mai Music composer does — the same flow works on any serious text to music platform.
- Describe the song in a brief — genre, BPM, key, mood, vocal style and arrangement. A structured brief gets a dramatically better result from the AI music model than a one-liner.
- Write the lyrics with section tags like [Verse], [Chorus] and [Bridge], or let the AI music generator write them for you from your brief.
- Set the length — from a few seconds to five minutes — and pick your music model: MiniMax Music 3 by default, Suno v5.5 when you need its wider toolkit.
- Hit create and check the credit cost shown on the button. The exact AI music generation price appears before you spend anything, and the finished track lands in your library.
The whole AI music generation takes about a minute for a short song. Because the model writes the arrangement in one pass, the vocal stays consistent from the first verse to the final chorus — which is what separates full-length AI music from thirty-second fragments.
MiniMax Music 3 vs Suno v5.5: which AI music model should you pick?
Mai Music runs two leading music models. MiniMax Music 3 is the default: it produces complete, structurally coherent songs up to five minutes long from a brief and lyrics in a single pass, with open weights published under MiniMax's community licence.
Suno v5.5 is the wider toolkit: reference audio, saved voices, stem separation, covers and extensions, with two takes per generation scored separately. Choose Suno when you need to remix, split or extend existing music rather than start from text.
Both are true AI music generators, and the composer's model selector switches between them instantly — the same brief and lyrics work on either engine.
Is there a free AI music generator?
Yes — Mai Music is a free AI music generator to start with. Signing up puts 15 credits in your account: enough for a full sixty-second song or three short sketches, with no card required. The exact credit cost of each generation is shown on the Create button before you spend anything.
Credits scale with duration, so a fifteen-second sketch costs a fraction of a five-minute track. Everything you generate stays in your library with its brief and lyrics attached, and every track downloads as a file.
FAQ
MiniMax Music 3 questions, answered
The questions people ask before their first song. If yours is not here, the contact page reaches a person.
Your first Mai Music song is about four sentences away
Write a brief, write or generate the words, pick a length. 15 credits land in your account when you sign up — enough to hear what MiniMax Music 3 does with an idea before you decide anything.