So, I says to myself, I says… Why toss coats into blender? Why not try making a LoRA based on one coat?

But the training data, you say, too thin, too homogeneous! Yes, solid points, but I flaunt my vibe-coded variety generator app, API key, and frontier image models at you. And the ten bucks I threw at them… Don’t mention Qwen, it’s… not great at inferring what fur looks like from different angles, and I’ll leave it at that.

This page will be updated with new coats as… I feel like it.

One important general note about all of them: they will be ‘less flexible’ than the generalist LoRAs. By that I mean they will try to create the exact coat they are trained on. If there’s something about the coat you don’t like, I’d skip that one. On all other things, they should be flexible: lighting and character for example.

Coat 2: Jacket-length blue fox.

Because sometimes less is more. Besides, options- we’re about the options. And even our options have options, because this one will require more careful prompting; otherwise you’re gonna just get a coat. If that sounds cool, then by all means, ignore the following:

Use Tips:

• The term you want is “fur jacket,” not “fur coat”. That’s the minimum spec; the recommended spec is “short fur jacket.” Because AI is dumb.

• Keyword “bbffj01” can help, but it's not necessary, as usual.

• In this case, tread carefully with your “size superlatives,” because some may push your jacket back into a coat. Grounding with “short fur jacket” is always a good start.

• Prompt some small, hand-related detail if you’re worried about the sleeves swallowing them up. (Again, you may be pumping your fist going “hell yeah, that sounds great!”). Nail polish, gloves, etc., they all work and can get those phalanges out of hiding (usually).

Coat 1: Full-length blue fox with notched lapels.

Usage Tips:

• The trigger is "1bbffcoat," but it'll kick in for any "fur coat" prompt. The prompt probably will change for future coats. I dropped ‘bluefoxfur’ because Krea was picking up the ‘blue’ in a few cases.

• Sampler/Scheduler: Notably dependent on the checkpoint. I’m now using Kroma for all previews, because I like Kroma. Start with whatever the model/checkpoint recommends and go from there. I would recommend using either a “detail-oriented” scheduler or sampler, not both; texture starts to get a little overbaked otherwise.

• Example: ER-EDE or EXP Huen 2 x0 SDE pairs with Simple. Or Euler with Beta_57.

• I generally get good results with Euler Ancestral and Beta_57.

• For basic “fashion” shots, 1 is fine. If you plan to have many body parts interacting, you may want to dial it down a bit.

• At a lower strength, it can be a good ‘texture booster’ to make a checkpoint’s fur look a little better without strongly overriding what it would have given you anyway.

• This one generally works well with specific coloring prompts, ex: “red dyed” or “coat colored red.” Though the standard checkpoint/sampler/scheduler disclaimer applies (it may not in every single combination of those ever).

• The coat will be “large” by default, but feel free to go nuts on your favorite adjectives: huge, oversized, voluminous. It should respond favorably. The sample prompts include ideas on how to ‘boost your volume’ when generating for less-than-reasonably-sized garments.