Minimax music3 音乐生成

基础模型 : MiniMax H3|发布时间 : 2026-09-17 15:33

DBclown（LH）

Music description

Global Metadata: Lo-fi hip-hop, chillhop. 78 BPM, D flat major, major scale with jazzy extensions. Laid-back and dreamy throughout, a gentle warm drift with a subtle late-night glow that deepens in the middle and dissolves softly at the end. Studying, raining-outside, headphones-on late-night listening. Bedroom production: muddy warm texture, heavy vinyl crackle, tape hiss and wow-flutter pitch wobble, low-passed dusty mix, soft-clipped drums, everything slightly detuned and cozy.

Vocal Details: Soft androgynous vocal, hushed half-sung half-spoken delivery, sitting low in the mix like another instrument, lazy behind-the-beat phrasing, gentle breathy timbre. Sparse murmured double-tracked harmonies, occasional wordless "mmm" and "ooh" hums drenched in tape delay and warm spring reverb. Long stretches with no vocals at all.

Arrangement: Dusty boom-bap drums with a soft thumping kick, cracked snare with lazy swing, brushed hi-hats, low round sub bass. Warm Rhodes piano chords with slow chorus wobble as the harmonic bed, mellow jazzy guitar licks answering the vocal lines, constant vinyl crackle as texture. Intro: rain and vinyl noise, solo Rhodes chords fading in, drums slipping in halfway. Verses: minimal — drums, bass, Rhodes, soft guitar fills between lines. Instrumental sections: guitar and Rhodes trade relaxed jazzy phrases over the beat, occasional muted trumpet ghost notes far in the background. Bridge: drums drop away to rain, crackle, and floating detuned Rhodes, then the beat eases back in. Outro: elements fade one by one until only vinyl crackle and a last unresolved Rhodes chord remain.

Lyrics

Max duration

使用指南

MiniMax Music 3是一款高性能音乐生成模型，可生成时长不超过五分钟的完整歌曲。它以歌词和详细的音乐描述为条件，生成结构连贯、人声富有表现力、编曲不断演进且音频质量稳定的长篇歌曲。

输入设置： 1、**Lyrics **：定义要演唱的词句，并可包含明确的段落标签，如 [Intro]（前奏）、[Verse]（主歌）、[Pre-Chorus]（预副歌）、[Chorus]（副歌）、[Post-Chorus]（后副歌）、[Bridge]（桥段）、[Instrumental]（器乐段）、[Solo]（独奏）和 [Outro]（尾奏）。 2、Music description：定义音乐风格、情感演进、人声表现、乐器配置、编曲方式及制作特性。 3、max duration：生成最大时长。

关于Music description建议使用包含三个部分的结构化标题：全局元数据：流派、子流派、BPM、调性、音阶、情感发展、聆听场景和制作概况。人声细节：人声性别、音色、演唱风格、和声、伴唱和人声效果。编曲：主要和次要乐器、乐器组级演变、节奏、低音、打击乐、织体和空间效果。这种表示方法使模型不仅可以遵循整体风格，还可以遵循歌曲随时间推移的音乐发展。

Minimax music3 音乐生成. MiniMax Music 3是一款高性能音乐生成模型，可生成时长不超过五分钟的完整歌曲。它以歌词和详细的音乐描述为条件，生成结构连贯、人声富有表现力、编曲不断演进且音频质量稳定的长篇歌曲。输入设置： 1、 Lyrics ：定义要演唱的词句，并可包含明确的段落标签，如 [Intro]（前奏）、[Verse]（主歌）…

作品

暂无作品