MiniMax Music 2.6: Four Stories We Want to Tell

Today we release MiniMax Music 2.6.

In past generations, we talked about how much stronger the model had become. This time, we want to tell it differently — through four people, four pieces of music, and four things that weren't possible before. Because a music model's upgrade doesn't live in spec sheets. It lives in "this time, someone used it to do something they couldn't do before."

She's Choreographing a Flamenco Solo in a Brooklyn Studio

@ElenaVargas is a flamenco dancer who teaches and performs in Brooklyn. She's building a solo piece for a spring showcase. Lights down, bare stage, just her shoes on wood and whatever's coming through the speakers. She needs a track she can actually move to, but the right one doesn't seem to exist. Stock libraries file flamenco under "world music," all smooth guitars and zero edge. Hiring a guitarist for every rehearsal isn't realistic.

The hardest thing about flamenco for AI isn't the notes. It's the silence between them.

The guitar accelerates into a rapid strum. Everything halts. Two beats of pure stillness heavy enough to hold the whole room. Then it explodes back. The handclaps don't land on the beat. They orbit around it, creating tension that pulls your body forward before the resolution hits. Previous AI played flamenco in time. Real flamenco plays with time.

Music 2.6 models the architecture of tension. Not just what's played, but when things stop and how they restart. In her track "Corazón de Fuego," the guitar builds through a shimmering passage then drops into a sudden stop so sharp the room forgets to breathe. The handclaps re-enter a half-beat ahead of the guitar, pulling the rhythm forward instead of following it. The drum answers with one dry slap that locks everything back in. Musicians watching a dancer, not reading a score.

What she can do now: tell 2.6 "flamenco, dramatic pauses, the silence matters more than the guitar." The rest is footwork.

He's Scoring a Boss Fight for His Indie Game

@BenMingYanZu is building an action game by himself. No team. No budget for a composer.

Until now, his choices were: drop thousands on a sample library and pray players don't hear the same loop twice, or use AI-generated music that could do "loud" but couldn't do "heavy." Drums with no chest punch. Low end that turned to mush the second you put on decent headphones. Nothing that could sell a boss fight.

Music 2.6 completely reworked its low end. The sub-bass actually hits now. Drums are tight, not boomy. On headphones, in a car, through a gaming rig, you feel it before you hear it.

And it follows dramatic direction. He can prompt "start oppressive, build toward power, erupt into invincibility" and the track actually builds that way.

What he can do now: score a full boss-fight soundtrack in an afternoon. No sample library. No composer. No five-figure invoice.

She's Looking for an Exclusive Playlist That Can Fill a Cafe for Four Hours

@MissNanFangYi runs a specialty coffee shop. Every day from 2 to 6 PM, she needs music playing. She's never once found the right playlist.

The "cafe music" playlists on streaming are either wallpaper — so smooth they disappear, might as well be silence — or main character energy — a sax solo so loud it kills the conversation two tables over. What a cafe actually needs sounds like a contradiction: good enough to notice, chill enough that nobody minds. Previous AI couldn't thread that needle.

Music 2.6 does something different with vocals and melody: it lets them be imperfect. In the right lane — lo-fi, indie folk, indie jazz — that imperfection is the vibe. It's what makes a track feel lived-in instead of generated.

Her track "Desert Race" sounds like someone singing to herself at 1 AM. Casual, a little defiant, not trying to impress anyone. The bass and drums carry as much weight as the voice, and the highs are rolled off into something warm and dark. Put it on in a cafe and nobody's annoyed. But every now and then, someone looks up and asks, "Who is this?"

What she can do now: skip the playlist rabbit hole. Just tell 2.6 "late-night feel, urban, slightly tipsy, not too bright." That's the whole brief.

She's a Daughter Making a Birthday Surprise for Her Mom

@NYX doesn't arrange music. Doesn't know any musicians. Doesn't have a big budget. She just wants to do one thing: take the song her mom used to play when she was little and redo it in her own style. A birthday surprise.

She couldn't do this before. Not because AI can't make music — it can. But she doesn't want a new song. She wants that song, just different. Same melody her mom would recognize from the first bar. New everything else. That's not generation. That's constraint.

This is what Cover does. It's new in Music 2.6.

Upload a track. The model pulls out the melodic skeleton and hands you the rest. Flip the genre from folk to metal. Swap a symphony for synths. Keep the melody, rewrite the lyrics. It's the same song and it's not.

"Auld Lang Syne" original:

Cover version:

What she can do now: the night before her mom's birthday, sit down for half an hour and make something that used to take a whole arranging team.

These Four People, and Everyone Else

Those were four stories. But Music 2.6's upgrades extend far beyond these four people —

First-packet latency reduced to under 20 seconds. After you finish writing your prompt, just one deep breath later you'll hear the first feedback. The "waiting" feeling is essentially gone.

Instruction control comprehensively enhanced. BPM, key, song structure, emotional arc — all can be written into your prompt and accurately executed by the model. The specific requirements you write down, the model takes seriously.

Mid-to-low frequency acoustics systematically optimized. Beyond the game scoring scenario mentioned earlier, any style that demands strong low end — House, Trap, Drum & Bass — will directly benefit.

These capabilities appeared in all four stories above, but their applicability extends far beyond those four.

Another Way to Use It: Let an Agent Do It for You

All four people above worked directly with Music 2.6. But if you're an AI Agent developer, you're probably not thinking about making a song yourself — you're thinking about how to let your Agent make a song.

That's what the three Music Skills we're open-sourcing alongside this release are designed to solve:

Try It Now

None of these four stories were scripted by Music 2.6. They were things these people wanted to do themselves — we just used AI to make it possible.

Starting today, MiniMax Music 2.6 opens its global creative beta — 14 days free, and you're invited to create.

Behind every piece of music is someone becoming a music creator for the first time in the age of AI.

What's the thing you've been wanting to do?

Product: www.minimax.io/audio/music

API: platform.minimax.io/docs/api-reference/music-generation

Intelligence with Everyone.