Providers and ModelsMiniMax

MiniMax Music 3

MiniMax Music 3
MiniMax · Audio Generation
POST /v1/audio/generations

Generates complete stereo songs from lyrics and a style prompt, with duration, seed, and format controls up to 5 minutes.

At a glance

FieldValue
Model idminimax-music-3
Model release date2026-08-13
Input modalitiesText
Output modalitiesAudio
Context window-
Weight precisionINT8
Featuresmusic_generation, lyrics, text_to_music, seed_control
Native inferenceYes
NewYes
Supported endpointsPOST /v1/audio/generations
Alternate model idsminimax-music3, minimax-music3.0, minimax/music3, minimax/minimax-music3, minimaxai/minimax-music3, MiniMax-Music3

Pricing

ChargeSpecRate
Music generationper generated second$0.0018 (was $0.002)

Example request

$curl https://api.empiriolabs.ai/v1/audio/generations \
>  -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
>  -H 'Content-Type: application/json' \
>  -d '{"model": "minimax-music-3", "prompt": "warm jazz piano", "duration": 8}'

Parameters

ParameterTypeRequiredDefaultDescription
promptstringyes-Style and arrangement instructions for MiniMax Music 3. Describe genre, mood, vocal character, and instrumentation.
lyricsstringno-Song lyrics. Required for a vocal track. Hidden when Instrumental is on. Put one structure tag on its own line before each section. Supported tags: [Intro], [Verse], [Pre Chorus], [Chorus], [Interlude], [Bridge], [Outro], [Post Chorus], [Transition], [Break], [Hook], [Build Up], [Inst], [Solo]. Lowercase is fine.
instrumentalbooleannotrueWhen on, the model writes an instrumental track and lyrics stay hidden. Turn this off to write a vocal song.
audio_durationnumberno10Length of the generated song in seconds. MiniMax Music 3 accepts 10 to 300 seconds and bills a 10 second minimum. · Range: 10 – 300
seedintegerno-Random seed for reproducibility. Same seed and identical params produce the same song.
formatenumno"wav"Audio container format. WAV is uncompressed, FLAC is lossless, MP3 is a smaller file. · Allowed: wav, flac, mp3
return_base64booleannofalseWhen true, the response includes the rendered audio as base64.
response_formatenumno"url"How the worker returns the audio. url returns a signed URL; b64_json inlines the bytes. · Allowed: url, b64_json

Notes

Defaults

  • Instrumental is on by default. Pass lyrics and set instrumental to false for a vocal track
  • 10 second songs, billed at a 10 second minimum
  • WAV output
  • Pinned at 30 steps and guidance 1.7

Controls

Supports prompt (style and arrangement), lyrics with structure tags, instrumental mode, 10-300s duration, seed, and WAV, FLAC, or MP3 output. URL or base64 response format.

Limits

  • Duration 10 to 300 seconds
  • Songs are billed per generated second with a 10 second floor

Billing

  • Charged per generated second at the catalog rate. A 10 second request that returns a 10 second song bills 10 seconds.
  • A request rejected before generation is not billed.

Machine-readable schema: GET https://api.empiriolabs.ai/v1/models/minimax-music-3.