* New lora for this uploaded with smaller file size and generally way better,its been uploaded separately due to reasons I put in that desc *
try using v2 its a lot better than this.
Description
I trained this on "bl4cked sports bra and thong" and forgot to add wearing because I'm special so sometimes separating the wearing from that phrase can help.
FAQ
Comments (17)
plays nice with character loras in my brief testing. will upload some images for reference.
800mb is brutal though. I'll see if I can generate anything at all.
@lost_moon I said sorry ! xD forgive me
make a bleached version
Unless im paid to do that its wayy to much work xD
@Gentlemanbunny just say you hate it, that you only like blacked and hate bleached. No need to censor yourself by using excuses like it's too much work.
Just be truthful👍
@Casanova_ I'd be all for it but I would have to somehow generate or edit 30+ picture added to a dataset and then make it decent enough quality I could then teach flux it. and thats too much work for my lazy ass.
I also made a Flux LORA for Blacked Clothing, but feel like it doesn't work very well. What was your data set, captions, settings, etc? I'm happy to share my dataset with you too if it would help you.
Hey, if you join pyro's discord im on there a lot discussing the ins and outs of it :)
@Gentlemanbunny do you have a link?
@BinaryBottleBake not on hand, just search for pyro on here he posted one the other day with his lora
@Gentlemanbunny Yeah, I tried the discord links from a few of his recent posts and they all say "invite invalid" or "link broken"
@BinaryBottleBake https://discord.gg/n9bTJAPD try that
Great work on this - seems relatively flexible and works well with character loras. I'm using 30 steps and seems to get the writing perfect about 30% of the time.
Thanks :) yeah I had to use a lower step version as I seemed to of trained the body as well but glad its working well flexibility is what I want most ! even if it takes a few more renders.
@Gentlemanbunny Done a bit more testing with it - you're right the body is trained in as well. That's okay for a v1 though, would love to see what a v2 looks like! Definitely prefers the front on angle but I've had some luck with other angles as well, just takes more rolls.
@jeremygin yeah for v2 I'm thinking I will try training it without a token, just try to override a concept flux already sorta knows ? that or maybe I need to caption the body as well who knows xD
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.