$5 free credits when you sign up Claim now
Marlin 2b now available Test it!
MiniMax H3 now available! Test it!
MiniMax Music 3 now available! Test it!
Browse all Models
Browse all
Browse all
Browse all
  • Text-to-Music

MiniMax Music 3

MiniMax Music 3 generates complete songs with sung vocals, up to five minutes, from a structured description and tagged lyrics. Open weights from MiniMax, running on deAPI's GPU network.

  • Free tier available
  • No subscriptions
  • No credit card required

Sample Outputs

Real outputs generated with MiniMax Music 3 on deAPI.

“Global Metadata Basic Attributes: bpm is 68. key is F, and scale is minor. Vocal Jazz / Noir Ballad. Global Emotional Progression: Late-night and unhurried. It begins lonely and resigned, warms slightly in the chorus, hands the feeling to a muted trumpet in the solo, and ends on a quiet, unresolved sigh. Sonics & Production Profile: A small club recorded live to tape. Warm, dark and roomy, with soft tape hiss, a close vocal microphone and a wide natural stereo image of the band. Vocal Detail…” “Global Metadata Basic Attributes: bpm is 68. key is F, and scale is minor. Vocal Jazz / Noir Ballad. Global Emotional Progression: Late-night and unhurried. It begins lonely and resigned, warms slightly in the chorus, hands the feeling to a muted trumpet in the solo, and ends on a quiet, unresolved sigh. Sonics & Production Profile: A small club recorded live to tape. Warm, dark and roomy, with soft tape hiss, a close vocal microphone and a wide natural stereo image of the band. Vocal Details Vocal Gender & Timbre: Female lead, smoky low alto with a velvety, slightly husky grain. Vocal Style: Behind-the-beat phrasing, soft consonants, long held notes that fade with slow vibrato, a near whisper at the ends of lines. Harmony/Backing Vocals: None. One voice only. Vocal FX: Light plate reverb, no other processing. Arrangement Primary: Felt-soft piano playing sparse minor ninth chords, and a walking upright bass. Secondary: Brushed snare and ride cymbal. A harmon-muted trumpet answers the voice and takes the solo. Groove & Foundation Progression: The intro is piano alone for four bars. Bass and brushes enter with the first verse in a slow swing. The chorus lifts a little with fuller piano voicings. The solo section is trumpet over bass and brushes with the piano comping quietly. The last chorus returns softer than the first, and the outro leaves piano and bass alone. Embellishments, Textures & Spatial FX: Trumpet fills between vocal lines, a soft cymbal swell into each chorus, room ambience and a final piano chord left to ring.”

“Global Metadata Basic Attributes: bpm is 126. key is A, and scale is minor. Progressive House / Melodic EDM. Global Emotional Progression: Starts weightless and hopeful, tightens with rising tension through the pre-chorus, releases into a euphoric drop, and settles into a calm afterglow. Sonics & Production Profile: Wide, glossy and modern festival production. Deep clean sub, bright airy top end, heavy sidechain pumping on the pads, long reverb tails. Vocal Details Vocal Gender & Timbre: Fe…” “Global Metadata Basic Attributes: bpm is 126. key is A, and scale is minor. Progressive House / Melodic EDM. Global Emotional Progression: Starts weightless and hopeful, tightens with rising tension through the pre-chorus, releases into a euphoric drop, and settles into a calm afterglow. Sonics & Production Profile: Wide, glossy and modern festival production. Deep clean sub, bright airy top end, heavy sidechain pumping on the pads, long reverb tails. Vocal Details Vocal Gender & Timbre: Female lead, airy and crystalline soprano. Vocal Style: Breathy and intimate in the verse, opening into long sustained notes in the chorus. Harmony/Backing Vocals: Stacked octave doubles and wide harmonies in the chorus. Chopped vocal syllables are used as a rhythmic instrument in the drop. Vocal FX: Long hall reverb, dotted eighth-note delay throws on line endings, light pitch-perfect tuning. Arrangement Primary: Plucked synth arpeggio and a supersaw chord stack. Secondary: Four-on-the-floor kick, offbeat open hi-hat, clap on two and four, rolling sub bass, white-noise risers. Groove & Foundation Progression: The intro is filtered pads and the pluck arpeggio with no drums. The verse adds a soft kick and sub under the vocal. The pre-chorus removes the kick, opens the filter and builds with a snare roll and riser. The chorus brings full drums and supersaws under the vocal. The instrumental section is the drop: lead synth melody, vocal chops and the hardest kick. The final chorus is the fullest, and the outro strips back to pads and arpeggio. Embellishments, Textures & Spatial FX: Reverse cymbals into each section, a one-beat silence before the drop, wide stereo delays, a filtered fade at the end.”

“Global Metadata Basic Attributes: bpm is 90. key is E, and scale is minor. Boom Bap / Jazz Rap. Global Emotional Progression: Calm, proud and warm from start to finish. The verses are focused storytelling and the hooks lift into a friendly neighbourhood chant. Sonics & Production Profile: Dusty nineties sampler aesthetic. Punchy mid-heavy drums, rounded bass, vinyl crackle, a lo-fi jazz piano loop, and a dry vocal sitting up front. Vocal Details Vocal Gender & Timbre: Male lead, relaxed mid…” “Global Metadata Basic Attributes: bpm is 90. key is E, and scale is minor. Boom Bap / Jazz Rap. Global Emotional Progression: Calm, proud and warm from start to finish. The verses are focused storytelling and the hooks lift into a friendly neighbourhood chant. Sonics & Production Profile: Dusty nineties sampler aesthetic. Punchy mid-heavy drums, rounded bass, vinyl crackle, a lo-fi jazz piano loop, and a dry vocal sitting up front. Vocal Details Vocal Gender & Timbre: Male lead, relaxed mid-low voice with a slightly raspy grain. Vocal Style: Laid-back conversational rap flow with clear diction, landing just behind the snare, emphasising end rhymes without shouting. Harmony/Backing Vocals: The hook is doubled and joined by a small group chanting the word "Rise". Occasional ad-libs sit low in the mix. Vocal FX: Mostly dry with light compression, and a short slap delay on the hook. Arrangement Primary: A looped, slightly detuned jazz piano chord sample and a swung drum break with a hard kick and cracking snare. Secondary: Warm upright-style bass line, vinyl crackle, a muted trumpet phrase, and turntable scratches. Groove & Foundation Progression: Two bars of piano loop and crackle open the track before the drums drop in with the first verse. The hook adds the trumpet phrase and group chant. The second verse removes the bass for the first two bars, then brings it back. The final hook adds scratches, and the outro is the piano loop fading with crackle. Embellishments, Textures & Spatial FX: Drum drop-outs on punchlines, a reversed cymbal into each hook, scratch fills between hook lines.”

“Global Metadata Basic Attributes: bpm is 108. key is E, and scale is minor. Spaghetti Western / Cinematic Instrumental. Global Emotional Progression: A lone, tense stand-off opening that breaks into a galloping ride, peaks with a proud trumpet theme, and rides off into silence. Sonics & Production Profile: Sixties film-score recording. Wide, dry and dusty, with a spring reverb on the guitar, a bright trumpet and a big natural room on the drums. Vocal Details The piece is instrumental. There…” “Global Metadata Basic Attributes: bpm is 108. key is E, and scale is minor. Spaghetti Western / Cinematic Instrumental. Global Emotional Progression: A lone, tense stand-off opening that breaks into a galloping ride, peaks with a proud trumpet theme, and rides off into silence. Sonics & Production Profile: Sixties film-score recording. Wide, dry and dusty, with a spring reverb on the guitar, a bright trumpet and a big natural room on the drums. Vocal Details The piece is instrumental. There are no sung or spoken words. A human whistle carries the main melody and a wordless male choir sings only the syllable "ah" in the final section. Arrangement Primary: A whistled melody and a twanging baritone electric guitar with tremolo and spring reverb. Secondary: Mariachi-style trumpet, nylon-string guitar strumming, galloping snare and tom pattern, tubular bell, jaw harp, castanets. Groove & Foundation Progression: The intro is a single bell, wind, and the baritone guitar playing the theme alone. The first instrumental section brings the whistle melody over a galloping drum and strummed nylon guitar. The solo section hands the theme to the trumpet, with the guitar answering each phrase. The final instrumental section stacks whistle, trumpet and wordless choir at full gallop. The outro drops to the lone whistle and a last bell. Embellishments, Textures & Spatial FX: Whip cracks, a jaw harp twang between phrases, castanet rolls, a long reverb tail on the last note.”

“Global Metadata Basic Attributes: bpm is 172. key is E, and scale is major. Pop Punk / Power Pop. Global Emotional Progression: Cheeky and restless in the verses, winding up through the pre-chorus and exploding into a joyful, shout-along chorus. The energy never drops and the song ends on a single shouted line. Sonics & Production Profile: Bright, loud and tight early-2000s rock production. Wall of doubled guitars panned hard left and right, punchy snare, compressed and upfront. Vocal Detail…” “Global Metadata Basic Attributes: bpm is 172. key is E, and scale is major. Pop Punk / Power Pop. Global Emotional Progression: Cheeky and restless in the verses, winding up through the pre-chorus and exploding into a joyful, shout-along chorus. The energy never drops and the song ends on a single shouted line. Sonics & Production Profile: Bright, loud and tight early-2000s rock production. Wall of doubled guitars panned hard left and right, punchy snare, compressed and upfront. Vocal Details Vocal Gender & Timbre: Male lead, bright nasal tenor with a youthful, slightly snotty edge. Vocal Style: Fast, punchy and melodic, clipped phrasing in the verses and wide open belted notes in the chorus. Harmony/Backing Vocals: Gang vocals shout "Monday is cancelled" with the lead in every chorus. High harmony a third above on the last line of each chorus. Vocal FX: Tight doubling on the lead, almost no reverb. Arrangement Primary: Distorted power-chord electric guitars and a fast punk drum beat. Secondary: Picked eighth-note bass, a palm-muted second guitar, hand claps and a tambourine. Groove & Foundation Progression: The first verse is palm-muted guitar, bass and a tight hi-hat beat. The pre-chorus stops everything for one bar except a snare roll. The chorus opens to full strummed chords, crash cymbals and gang vocals. Verse two adds a lead guitar octave line. The last chorus is the biggest, with claps on every beat, and the outro is one final hit and a shout. Embellishments, Textures & Spatial FX: Guitar feedback before the first chord, a pick slide into each chorus, drum fills at every transition.”

“Global Metadata Basic Attributes: bpm is 112. key is F sharp, and scale is major. Japanese City Pop / Eighties Funk Pop. Sung in Japanese. Global Emotional Progression: Breezy, nostalgic and bittersweet. The verses are cool and reflective, the chorus opens up bright and wide like a night drive along the coast, and the ending fades with a wistful smile. Sonics & Production Profile: Glossy early-eighties Tokyo studio sound. Clean, bright and spacious, with chorus-effect guitars, a punchy gated…” “Global Metadata Basic Attributes: bpm is 112. key is F sharp, and scale is major. Japanese City Pop / Eighties Funk Pop. Sung in Japanese. Global Emotional Progression: Breezy, nostalgic and bittersweet. The verses are cool and reflective, the chorus opens up bright and wide like a night drive along the coast, and the ending fades with a wistful smile. Sonics & Production Profile: Glossy early-eighties Tokyo studio sound. Clean, bright and spacious, with chorus-effect guitars, a punchy gated snare, glassy electric piano and a warm analog sheen. Vocal Details Vocal Gender & Timbre: Female lead, clear, sweet and bright with a light, youthful tone. Vocal Style: Smooth and precise with gentle vibrato at the ends of phrases, relaxed in the verses and soaring in the chorus. Harmony/Backing Vocals: Soft female harmony stacks in the chorus and airy "ooh" pads under the second verse. Vocal FX: Bright plate reverb and a short stereo delay. Arrangement Primary: Electric piano playing rich major seventh chords and a melodic slap bass. Secondary: Clean electric guitar with chorus effect playing funky single-note lines, a tight drum kit with gated reverb snare, analog brass synth stabs, tenor saxophone, shimmering synth bells. Groove & Foundation Progression: The first verse rides on electric piano, bass and a tight drum groove. The chorus adds brass synth stabs, guitar chords and harmony vocals. The second verse brings in a guitar counter-line. The final chorus adds a saxophone answering the vocal, and the outro lets the saxophone play over the fading groove. Embellishments, Textures & Spatial FX: A synth bell glissando into each chorus, tom fills, a wind chime at the very start.”

Capabilities

What you can do with MiniMax Music 3.

Text-to-Music

Generate original music from text prompts. Create background tracks, jingles, and full compositions with customizable duration.

Technical Specifications

Inference Steps

30 steps

Duration

20–300 seconds

How to use in deAPI

Integrate MiniMax Music 3 into your app with a single API call.

POST https://api.deapi.ai/api/v2/audio/music

curl · deAPI audio/music

curl -X 'POST' \
  'https://api.deapi.ai/api/v2/audio/music' \
  -H 'accept: application/json' \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "MiniMaxMusic3_11B_INT8",
    "caption": "upbeat electronic background music",
    "lyrics": "[Instrumental]",
    "duration": 20,
    "inference_steps": 30,
    "guidance_scale": 1.7,
    "seed": -1,
    "format": "mp3"
}'{
    "model": "MiniMaxMusic3_11B_INT8",
    "caption": "upbeat electronic background music",
    "lyrics": "[Instrumental]",
    "duration": 20,
    "inference_steps": 30,
    "guidance_scale": 1.7,
    "seed": -1,
    "format": "mp3"
}

Tip: The API returns a request_id. Use webhooks (recommended) or poll GET /jobs/{request_id} for results.

Need an API key?

Sign up for free and get $5 in credits to start.

Get $5 credits

Frequently Asked Questions

An open-weights music generation model released by MiniMax in August 2026. It produces a complete song with sung vocals in one pass, up to five minutes long, from a text description and lyrics.
Two text fields. The caption describes the sound: genre, tempo, key, mood, the voice and the arrangement. The lyrics hold the words to be sung, with a section tag such as [Verse] or [Chorus] on its own line above each section. Steps and guidance are fixed at 30 and 1.7, the values the weights were tuned for, and still have to be sent with exactly those values.
MiniMax recommends a structured caption in three parts: Global Metadata (genre, bpm, key and scale, emotional progression, production character), Vocal Details (gender, timbre, delivery, backing vocals, effects) and Arrangement (which instruments play and what enters, leaves or builds in each section). A one-sentence description works too and still produces a complete song, with less control over the result. See the prompting guide for a template.
Nine: [Intro], [Verse], [Pre-Chorus], [Chorus], [Post-Chorus], [Bridge], [Instrumental], [Solo] and [Outro]. Put each on its own line. Write the chorus words out each time you want them sung. A tag with no lines under it becomes a passage without singing.
In the caption, for example "bpm is 92. key is D, and scale is minor." The model has no separate inputs for tempo, key or meter, so leave the bpm field out. MiniMax notes that these are followed as guidance and may not match exactly.
Yes. Send lyrics with section tags and no words, for example [Intro], [Instrumental] and [Outro] on separate lines, and name the instrument that carries the melody in the caption. The track ends on its own. Sending only "[Instrumental]" also gives a track without vocals, but it fills the whole duration you request and may stop abruptly at the end.
Yes. The same caption, lyrics, duration and seed return an identical file. Changing the duration changes the song, even with the same seed.
The weights are released under the MiniMax-Music3 Community License, which permits commercial use with conditions, including visible attribution to MiniMax-Music3 in commercial products and separate authorization from MiniMax for businesses above $20 million in annual revenue. Read the license before shipping.
Open the Text-to-Music playground and select MiniMax Music 3, or send "model": "MiniMaxMusic3_11B_INT8" to POST /api/v2/audio/music. See https://docs.deapi.ai for code examples.
  • No subscription
  • No credit card required
  • Pay only for what you use

Try MiniMax Music 3 — $5 for free

Get $5 in free credits and start generating with MiniMax Music 3.