Update: Please if you see any of my models somewhere else report or let me know! Thnx
Trained on base model xl 1.0 + F1D + zit
Please, check version info. "about this version" box on right.
Feedback and ratings are much appreciated.
Gun Models xl:
Gun Models F1D:
Description
This is v1 of gun style for F1D.
It's important to understand this version doesn't understand gun models. But it understand basic guns. Like for example, Shotgun, pistol, revolver, rifle, etc.
It may understand famous weapons, like ak47, beretta, colt, etc. (Though it may struggle)
Hopefully, in next update(s) things might get better.
FAQ
Comments (13)
Yo! so you know how Negative_hand is a thing? make that but Gun we NEED Negative_Gun!
Requesting a M134 minigun please.
no keywords on flux? 😮
(I checked it in notepad++)
It would be appropriate if it had them, in one of the photos generated in your showcase you asked for an mp5 and it generated AR/M16, and M82 barret same generated M16, Desert Eagle is not a desert eagle and FN P90 generated kinda MP5k
With a single keyword in each photo of the dataset corresponding to each weapon, it would be perfect. And those with a suppressor silencer, for example, "m16 with silencer", "uzi with silencer"
postscript: I just read the "about this version". What needs to be improved is only this.
I did mention this lora can only generate basic guns.
I did train it on all previous guns datasets with names, types, etc.
It didn't understand anything.
That's why I did mention this lora can only generate basic guns.
** I keep experimenting different type(s) of training till getting a better one.
@EauDeNoire Flex does much much better with captioning as opposed to tags. JoyCaption is great at captioning images for it
@EauDeNoire I suggest, for example, not using the word "weapon" in any of the weapons .txt, because every time someone uses weapon in a prompt, it will generate a mix of all the weapons.
In the LoRA that I did of the Renault Fuego I avoided mentioning that the Renault Fuego is a car (in tag and in caption, i used both), so that not all cars are Renault Fuego (although it is helpful to mention that the Renault Fuego is a car when you do the prompt to generate the image) but all the cars that are not Renault Fuego are a car.
It's also like doing a person's LoRA without using the word men or woman, so they can be combined with another person without them blending together more effectively. In my dog and cat's LoRA I also avoided using the word "animal", because that is what they have in common. With the same lora i can generate my dog alone or my cat alone without mix, and both in same generation with great chances to no mix
@vinque I read this comment awhile ago and planned on applying this strategy, I finally got around to doing it and it worked perfectly. If I could tip this comment, I would. Thanks.
@vinque Smart Strategy
Does not work with huggingface SDXL pipeline
"Gun Style" model is "Flux1D" not XL
....M 1911 A-1?
Very nice
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.