Model Auto-Download / Check

The MiniMax Music Production Toolkit example workflow quietly depends on more model files than any one person wants to track: the MiniMax Music 3 dit/text-encoder/VAE trio, three FlashSR weight files, a FLUX.2 Klein set for the cover art, and an LLM GGUF. MiniMaxModelAutodownload is the node that answers the obvious first question - do I actually have all of these? - before you burn an hour discovering you don't.

Drop it early in the workflow and it checks every file the workflow references, downloads the ones that have a URL configured, and logs the whole thing. You get a text report output describing each file's status, which you can feed anywhere you want a status string to live.

What it checks

Each category is its own toggle, all on by default:

• minimax_models - the MiniMax Music 3 generation files (dit, text encoder, VAE).
• flux2_models - the FLUX.2 Klein cover-art branch (dit, text encoder, VAE).
• flashsr_models - the three FlashSR weight files used by the integrated audio super-resolution node.
• llm_model - the example LLM GGUF the workflow references.
• auto_download - master switch for the downloading itself.

The node reads a declarative models_config.json in the toolkit folder, which maps each file to its target folder and, where one exists, a download URL.

The honest catch: it can only download some of them

This is where the "MiniMax" branding can mislead you into expecting a one-click setup. Look at what actually has a URL in models_config.json and it's mostly just the FlashSR weights - three files fetched from the jakeoneijk/FlashSR_weights HuggingFace dataset into models/audio/flashsr. The MiniMax Music 3 and FLUX.2 Klein files have empty URLs: they're gated/official-channel downloads with no public link, so the node reports them as missing with guidance instead of failing. Same for the example LLM GGUF unless you add a URL yourself.

So treat this node as a pre-flight checklist plus a helper for the auto-downloadable bits, not as a magic installer that makes the whole workflow appear. With auto_download ON, genuinely missing files without a URL are reported and the run continues; only an actual download failure raises an error. Set auto_download OFF and you get fail-fast behavior instead - useful when you want a hard stop the moment something's absent.

Installing

cd ComfyUI/custom_nodes
git clone https://github.com/jplenio/ComfyUI-MiniMax-Music-Production-Toolkit.git
cd ComfyUI-MiniMax-Music-Production-Toolkit
python -m pip install -r requirements.txt

Restart ComfyUI, hard-refresh the browser, and you'll find it under MiniMax Music Production Toolkit/utilities.

Two gotchas worth knowing. First, model paths resolve through ComfyUI's own folder_paths.models_dir - if you run ComfyUI with --models-directory, the node looks under that directory, not the default. The pack ships scripts/check_model_paths.py precisely so you can verify the resolution before pulling your hair out. Second, you don't actually need this node for the FlashSR and LLM parts of the chain: those nodes do their own lazy first-use downloads. This node's real value is the one-shot overview and the batch-friendly pre-check before a long run of generations.