Open-source end-to-end production toolkit for MiniMax Music 3

[jplenio]

I built an open-source end-to-end production toolkit for MiniMax Music 3 in ComfyUI, and v2.1 is out now. :)

The idea is simple:

Fill in a few fields → get a complete song production.

Choose things like genre, tempo, time signature, key, vocals, language, mood and length. A local GGUF LLM then creates the MiniMax caption, lyrics, title and cover-art prompt.

After MiniMax Music 3 generates the track, the workflow continues automatically:

→ De-clipping / overload repair
→ FlashSR audio super-resolution
→ High-frequency / cymbal shimmer repair
→ LUFS & true-peak release preparation
→ FLUX.2 cover artwork
→ FLAC + MP3 + metadata + production JSON

The toolkit now includes 239 curated genre templates across 31 categories, Custom Mode for completely free prompting, and v2.1 adds a full editable System Prompt section with different production-focused prompt profiles.

There is also a separate Audio Enhancement Lab workflow if you only want to process an existing track.

Everything is open source under MIT and designed for local ComfyUI setups.

🎧 35-track demo: demo link
💻 GitHub: project link

I’d especially be interested in feedback from people already using MiniMax Music 3: What part of the Music 3 workflow would you like to see improved next?