MiniMax CN T2I Config

MiniMax is the Chinese AI lab best known for the Hailuo video line, and its hosted image API - image-01 and image-01-live - is what MiniMaxCNT2IConfig wraps. Of all the image config nodes in MediaHub, this one has the most knobs: aspect ratios, batch count, style presets, prompt optimization, even custom output dimensions. If you want to see what a provider config node looks like when the provider actually has a rich API, this is it.

Like every config node in the pack, it outputs a MEDIAHUB_IMAGE_REQUEST, not an image. Wire that request into MediaHubGenerateImage alongside your key and that's where the API call and the bill happen.

The inputs that matter

• model - image-01 or image-01-live. The live variant is the faster/cheaper path; image-01 is the quality default. If you're iterating, start live, commit with the base.
• prompt - the text. This is also where prompt_optimizer earns its keep: when enabled, the provider rewrites/expands your prompt server-side before generating. It defaults to off - flip it on when your raw prompts come back undercooked.
• aspect_ratio - eight options from 1:1 to 21:9. Or pick custom in size_mode and set width/height yourself (512–2048, steps of 8) if you need a specific pixel size the ratios don't cover.
• n - how many images per request, 1–9. This multiplies your cost, so treat it as a budget lever, not a feature.
• use_seed / seed - reproducibility switch, off by default like most providers here.
• use_style + style_type + style_weight - the distinctive bit. use_style off by default; flip it on and pick a style preset (漫画 manga, 元气 "genki"/vibrant, 中世纪 medieval, 水彩 watercolor) and a style_weight from 0.01–1.0 (default 0.8) to control how hard that style is applied. This is a genuinely useful way to push a consistent look without prompt gymnastics.
• aigc_watermark - an AI-generated-content watermark flag, off by default.
• response_format - url or base64.

Output: request → MediaHubGenerateImage.

Why you'd reach for it

MiniMax's image API is one of the more capable hosted options, and this node is the easiest way to play with its odd corners - the style presets alone are worth a weekend. It's also the pack's best example of how a good config node turns a provider's messy parameter surface into ComfyUI widgets you can actually reason about.

Installing it

Standard MediaHub install:

cd ComfyUI/custom_nodes
git clone https://github.com/vantang/ComfyUI-MediaHub.git
/path/to/ComfyUI/python -m pip install -r ComfyUI-MediaHub/requirements.txt

Restart ComfyUI, find it under MediaHub/Provider/MiniMax CN/Image. Dependencies are requests, Pillow, numpy - nothing heavy. You need a MiniMax CN API key (note the "CN" - that's the mainland-China endpoint, api.minimaxi.com; it's a different console/account from MiniMax Global).

Common gotchas

The style_type values are Chinese by design and the config passes them through verbatim - don't "translate" them to English or the provider won't recognize them. Also remember n and higher resolution both cost more; a n=4 run with custom 2048×2048 is a different price tier than one 1024 image. And as with every node here: config alone generates nothing. Request → MediaHubGenerateImage → Preview Image, or you'll wonder where your image went.