♂️ Male Outfit

Male Outfit (class MaleOutfitNode, in the 👗 Outfit/Male menu) is exactly what it sounds like: the comfyui-outfit pack's prompt builder, tuned for a male subject. It's not a separate app - both outfit nodes are generated by the same factory function in dynamic_outfit_node.py and differ only in which part catalogs and presets they load from data/outfit/male/ and data/presets.json. If you've read the Female Outfit article, you know the drill; this is the men's rack.

You reach for this one the same way: pick a preset (the male list runs 28 deep and is a lot more fun than the female one - Arabian Nights, Detective Noir, Chef Master, Japanese Samurai, Cyberpunk Rebel, Indian Wedding), leave the part dropdowns on random, and let the seed assemble a coherent outfit instead of you describing a tuxedo in tags.

What's actually different

The part catalog is male-specific, which matters for how you use the randomizer:

• facial_hair is a male-only dropdown with 22 options, from clean shaven to full beard and everything between (fu manchu, balbo, friendly chops, five o'clock shadow...). This one has real prompt logic behind it: the builder auto-injects negatives so a clean-shaven pick doesn't come back with a beard - and vice versa. If you set facial_hair: clean shaven, expect "beard, mustache, goatee, stubble" to appear in the negative_prompt output.
• Male hairstyle and body_type lists replace the female ones - buzz cuts and fades instead of Bantu knots and bouffants, broad-shouldered and burly instead of hourglass and petite. The hairstyle choices also trigger the auto-negative logic (pick buzz cut and you won't get a ponytail).
• No cheeks/lips makeup parts here, though the hidden makeup_data plumbing still exists in case you load a female-style workflow by accident.

Everything else - the seed system, presets, outputs - matches the female node. Same four outputs: positive_prompt (to your text encoder), seed (to your KSampler for reproducibility), negative_prompt, and metadata (JSON of what it picked). Same style_seed trick: set your own value to vary the outfit while the image seed holds, which is the fastest way to brute-force "same guy, different look" batches.

The few inputs you'll set

• preset - none, random, or one of the 28 named presets. Presets fill only fields still set to none/random/empty, so anything you choose manually overrides.
• avoid_terms - comma-separated stuff you never want (e.g. tie, formal wear when you're fishing for casual fits). Folded into the negative prompt plus the pack's baseline negatives (blurry, low-res, bad anatomy, extra fingers...).
• seed / seed_mode / style_seed - the same four seed modes; style_seed is the one people underuse.

Install and gotchas

Identical to the rest of the pack: ComfyUI Manager → search "ComfyUI Outfit Selection", or

cd ComfyUI/custom_nodes/
git clone https://github.com/manifestations/comfyui-outfit.git

then restart. No model downloads for this node; deps are just requests, pillow, numpy, torch. The known gotchas are shared too: enable_cache (on by default) can make it look like an input change did nothing, and a preset with keys that no longer exist logs contains unknown keys and silently skips them. If your male subject is coming out clean-shaven when you asked for stubble, check the auto-negatives in the metadata output before blaming the model - the node may have decided your choice needed reinforcing.