MusicGPT is a free, open-source application that runs Meta's MusicGen AI model locally on your computer — no subscription, no cloud processing, no internet required after setup. This guide covers everything you need to install it, generate audio that is actually usable, write prompts with precision, integrate the output into your DAW, and understand exactly what the non-commercial license means for your workflow. From first launch to batch scripting a complete music library, this is the professional perspective.
Complete installation walkthrough for macOS (Apple Silicon and Intel), Windows, and Linux. Homebrew one-liner, binary download, Docker for CUDA GPU users — including model storage locations per platform, first-launch model download, and CLI confirmation. No Python, no machine learning frameworks, no developer tools required.
The six-element prompt framework (genre, tempo, instrumentation, mood, texture, ambience), specificity principles, genre-by-genre vocabulary, the three-stage generation workflow (small → medium → large), and building a personal prompt library. Know which words produce usable results — and which waste generation time.
The CC-BY-NC-4.0 licensing reality explained plainly: what non-commercial means, what is clearly permitted, what is clearly prohibited, and how to structure an MWHQ production workflow that keeps MusicGPT in its correct lane alongside commercially licensed tools. No legal gray areas unexplained.
Chapter 1 establishes exactly what MusicGPT is — the open-source gabotechs application running Meta's MusicGen model locally — and how it differs from the commercial musicgpt.com platform. Chapter 2 covers installation on every major platform. Chapters 3 and 4 walk through UI mode (the chat interface at localhost:8642) and CLI mode (direct terminal generation and batch scripting). Chapter 5 covers model size selection: when to use small, medium, large, and melody, and the three-stage generation workflow that eliminates wasted large-model generation time.
Chapters 6 and 7 are the prompt engineering core: the six-element framework, specificity principles, and advanced techniques including genre blending and production quality descriptors. Chapter 8 covers DAW integration: sample rate conversion, processing chains, sidechain compression for video integration. Chapter 9 is the complete licensing breakdown — CC-BY-NC-4.0 defined, permitted uses enumerated, prohibited uses enumerated, and the MWHQ workflow guidance that keeps AI music production legally sound.
Sir Leslie Jennings is the founder of M.U.S.I.C. World HQ and has spent three decades as a music technology reviewer, curriculum developer, and professional producer. This guide draws on direct MusicGPT production experience — the prompt formulations that work, the model size workflow that saves hours, and the processing chain that makes 32kHz AI audio sit cleanly in a professional mix.
The MWHQ approach: offer knowledge, not a workshop. Every chapter gives you something that makes every future MusicGPT session better — not a trick for today, a principle for life. The 35-item appendix is a permanent desk reference covering every command, every model benchmark, every storage directory path, and every key licensing point.
$19.99 — one-time purchase with immediate PDF download.
Use the Pro Workshop Discount at checkout to reduce the price to $14.99.
Use the All Access Workshop Discount at checkout to reduce the price to $9.99.
MusicGPT costs nothing. The gap between installing it and using it well is knowledge — specifically, knowledge of which prompts work, which model to use when, how to process the output for professional contexts, and where the non-commercial license boundary sits. This guide closes that gap. Read it at the terminal, generate alongside every chapter, and finish with a working workflow and a prompt library that produces usable audio every session.
