Skip to content
Back to search

minimax-music3

MiniMax Music 3 is a high-performance music generation model for creating complete songs up to five minutes long. Conditioned on lyrics and a detailed music description, it generates structurally…

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/minimax-ai-minimax-music3/badge.svg)](/listings/minimax-ai-minimax-music3)

HTML

<a href="/listings/minimax-ai-minimax-music3"><img src="/api/listings/minimax-ai-minimax-music3/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 157 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <div align="center">
  2. <img width="100%" src="figures/Music3.png" alt="MiniMax">
  3. </div>
  4. <p align="center">
  5. <a href="https://agent.minimax.io/" target="_blank"><img src="https://img.shields.io/badge/MiniMax%20Code-FF6C37?logo=minimax&logoColor=white" alt="MiniMax Code"></a>
  6. <a href="https://platform.minimax.io/docs/guides/text-generation" target="_blank"><img src="https://img.shields.io/badge/API-FF6C37?logo=minimax&logoColor=white" alt="API"></a>
  7. <a href="https://minimax-ai.github.io/music3-demo/" target="_blank"><img src="https://img.shields.io/badge/Live%20Demo-FF6C37?logo=githubpages&logoColor=white" alt="MiniMax Music 3 Live Demo"></a>
  8. <a href="https://www.minimax.io" target="_blank"><img src="https://img.shields.io/badge/MiniMax%20Website-FF6C37?logo=minimax&logoColor=white" alt="MiniMax Website"></a>
  9. <br>
  10. <a href="https://modelscope.cn/organization/minimax" target="_blank" rel="noopener noreferrer"><img alt="ModelScope MiniMax AI" src="https://img.shields.io/badge/ModelScope-MiniMax%20AI-white?labelColor=%23EF3D5D"></a>
  11. <a href="https://platform.minimaxi.com/docs/faq/contact-us" target="_blank"><img src="https://img.shields.io/badge/WeChat-07C160?logo=wechat&logoColor=white" alt="WeChat"></a>
  12. <a href="https://discord.com/invite/DPC4AHFCBw" target="_blank"><img src="https://img.shields.io/badge/Discord-5865F2?logo=discord&logoColor=white" alt="Discord"></a>
  13. <a href="https://huggingface.co/MiniMaxAI" target="_blank"><img src="https://img.shields.io/badge/Hugging%20Face-FFD21E?logo=huggingface&logoColor=black" alt="Hugging Face"></a>
  14. <a href="https://github.com/MiniMax-AI/MiniMax-Music3" target="_blank"><img src="https://img.shields.io/badge/GitHub-181717?logo=github&logoColor=white" alt="GitHub"></a>
  15. <a href="https://huggingface.co/MiniMaxAI/MiniMax-Music3/blob/main/LICENSE" target="_blank"><img src="https://img.shields.io/badge/LICENSE-4CAF50?logo=creativecommons&logoColor=white" alt="LICENSE"></a>
  16. </p>
  17. # MiniMax Music 3
  18. **MiniMax Music 3** is a high-performance music generation model for creating complete songs up to **five minutes** long. Conditioned on lyrics and a detailed music description, it generates structurally coherent songs with expressive vocals, evolving arrangements, and stable long-form audio quality.
  19. MiniMax Music 3 combines an **8B Global LLM** for long-range musical structure, a **0.6B Local LLM** for frame-level acoustic detail, and a continuous hidden-state synthesis system based on **Flow Matching** and **Flow-VAE**. The model produces 32 kHz, 16-bit stereo WAV audio.
  20. ## Prompt Enhancement
  21. A concise natural-language description can be used directly. For richer prompts and more precise control, use the provided [`music-caption-rewriter`](skills/music-caption-rewriter/SKILL.md) skill to expand it into a Structured Caption containing `Global Metadata`, `Vocal Details`, and `Arrangement`. The skill preserves musical instructions attached to lyric section tags in the arrangement description while keeping the lyric text in the lyrics input. See the [skills page](skills/README.md) for installation instructions and a full feature overview.
  22. ```bash
  23. npx skills add MiniMax-AI/MiniMax-Music3 --skill music-caption-rewriter
  24. ```
  25. ## Demo
  26. Explore music generation examples on the [MiniMax Music 3 Demo](https://minimax-ai.github.io/music3-demo/).
  27. <p align="center">
  28. <img width="100%" src="figures/music3.0-Architecture-Diagram.png" alt="MiniMax Music 3 architecture diagram">
  29. </p>
  30. ## Complete Songs with Long-Range Coherence
  31. MiniMax Music 3 natively supports full-song generation up to five minutes. The model maintains musical themes, rhythm, vocal identity, and arrangement progression across long sequences, enabling complete structures such as intro, verse, pre-chorus, chorus, bridge, instrumental break, and outro.
  32. ## Fine-Grained Music Control
  33. The model accepts two complementary inputs:
  34. - **Lyrics** define the words to be sung and may include explicit section tags such as `[Intro]`, `[Verse]`, `[Pre-Chorus]`, `[Chorus]`, `[Post-Chorus]`, `[Bridge]`, `[Instrumental]`, `[Solo]`, and `[Outro]`.
  35. - **Music description** defines the musical style, emotional progression, vocal performance, instrumentation, arrangement, and production profile.
  36. For precise control, we recommend using a Structured Caption with three sections:
  37. - **Global Metadata**: genre, subgenre, BPM, key, scale, emotional progression, listening scenario, and production profile.
  38. - **Vocal Details**: vocal gender, timbre, performance style, harmony, backing vocals, and vocal effects.
  39. - **Arrangement**: primary and secondary instruments, section-level instrument evolution, groove, bass, percussion, textures, and spatial effects.
  40. This representation allows the model to follow not only a global style, but also the musical development of the song over time.
  41. ## Hybrid-LM
  42. MiniMax Music 3 uses a hierarchical autoregressive architecture that separates global musical modeling from local acoustic modeling.
  43. - The **Global LLM (8B)** predicts the first RVQ codebook frame by frame and models the song's long-range semantic and structural progression.
  44. - The **Local LLM (0.6B)** predicts the remaining acoustic codebooks within each frame and restores fine-grained acoustic information.
  45. The Global LLM is initialized from Qwen3-8B. During training, its embedding and output layers are first adapted to semantic music tokens. The Global and Local LLMs are then jointly trained to model all RVQ codebooks.
  46. ## Continuous Hidden-State Synthesis
  47. Instead of decoding only from discrete RVQ tokens, the synthesis module fuses the final hidden states of the Global and Local LLMs. These continuous representations preserve richer acoustic information for vocal articulation, instrumental texture, and temporal continuity.
  48. The synthesis path is:
  49. ```text
  50. Global and Local LLM hidden states
  51. ↓
  52. Hidden-state fusion
  53. ↓
  54. Flow Matching (2.4B)
  55. ↓
  56. Flow-VAE latent
  57. ↓
  58. Flow-VAE Decoder (123M)
  59. ↓
  60. 32 kHz stereo audio
  61. ```
  62. The Flow-VAE architecture is adapted from MiniMax Speech and retrained for the dynamic range and spectral characteristics of music.
  63. ## Music Tokenizer
  64. The training tokenizer uses eight layers of Residual Vector Quantization (RVQ):
  65. - The first semantic codebook contains **16,384** entries and captures the core musical semantics and structure.
  66. - The remaining seven acoustic codebooks contain **1,024** entries each and represent residual acoustic details.
  67. Training first optimizes the semantic codebook, then jointly trains all eight codebooks. At inference time, waveform synthesis uses the fused LLM hidden states and does not require the discrete tokenizer decoder.
  68. ## How to Use
  69. ### Download the Model
  70. ```bash
  71. hf download MiniMaxAI/MiniMax-Music3 --local-dir /path/to/minimax_ttm
  72. ```
  73. ### Serve with SGLang-Omni
  74. MiniMax Music 3 is currently supported by [SGLang-Omni](https://github.com/sgl-project/sglang-omni). From the SGLang-Omni repository root, run:
  75. ```bash
  76. sgl-omni serve --model-path MiniMaxAI/MiniMax-Music3 --port 8000
  77. ```
  78. Inference uses two CUDA GPUs:
  79. - **GPU 0** runs Qwen3 and eight-codebook RVQ autoregressive generation.
  80. - **GPU 1** runs Flow Matching and DAV waveform decoding.
  81. ### Generate Music
  82. The service uses the shared speech API. Put the lyrics in `input` and the music description in `instructions`.
  83. ```bash
  84. curl http://127.0.0.1:8000/v1/audio/speech \
  85. -H 'Content-Type: application/json' \
  86. -d '{
  87. "model": "minimax_ttm",
  88. "input": "[Verse]\nMorning light filtering through the pine\n[Chorus]\nSoftly the world begins to breathe",
  89. "instructions": "A warm acoustic pop song with intimate female vocals, fingerpicked guitar, soft piano, and a gradual emotional build into a wide final chorus.",
  90. "response_format": "wav",
  91. "seed": 7,
  92. "max_new_tokens": 9000,
  93. "stream": false
  94. }' \
  95. --output minimax_music3.wav
  96. ```
  97. An example generated with this checkpoint is available at [`assets/minimax_ttm.wav`](assets/minimax_ttm.wav).
  98. ### Music Caption Rewriter Skill
  99. This repository includes `music-caption-rewriter`, an agent skill that turns a brief music description and optional tagged lyrics into a detailed Music 3.0 Structured Caption. It uses a bundled text library to select compatible style references, combine requested influences, and build a coherent section-by-section arrangement without external APIs or runtime dependencies.
  100. Then invoke `$music-caption-rewriter` in a supported coding agent with a music description and, optionally, lyrics containing section tags such as `[Verse]` or `[Chorus]`. The skill returns `Global Metadata`, `Vocal Details`, and `Arrangement` while preserving explicit musical constraints and keeping lyric text out of the rewritten caption.
  101. ## Limitations
  102. - Inference requires two CUDA GPUs.
  103. - Only non-streaming generation is currently supported.
  104. - The tokenized text prompt is limited to 5,000 tokens.
  105. - Audio generation is limited to 9,000 acoustic frames.
  106. - Section tags and music descriptions provide generative control rather than strict symbolic guarantees. The generated tempo, key, instrumentation, lyrics, and song structure may not always match every requested detail exactly.
  107. ## Contact Us
  108. Contact us at [model@minimax.io](mailto:model@minimax.io).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.