Hey everyone,
So, let's get straight to it.
I started making LoRAs for one reason:
To see you guys unleash your creativity and make killer images!
And honestly?
The stuff most of you come up with using my LoRAs has been next-level, seriously inspiring, and way beyond anything I could've cooked up myself. Major props for that.
When I put my LoRAs on Civitai, it was never about chasing cash.
I don’t care about the buzz, and I've never cashed out a dime.
Early Access?
That was purely to bankroll new LoRA training, simple as that.
Any Buzz I don't need usually goes to tipping cool art.
But here’s the flip side.
Over the last months, the game changed.
I’m seeing my distinct styles, my concepts, and even images made from my LoRAs get straight-up copied.
It happens on a month basis now.
People are training their own LoRAs on my stuff, sometimes churning out weak 1:1 knock-offs.
Last one I saw was just uploaded today.
Civitai’s hands-off approach here? - Sadly no help as they tolerate it.
I quote a staff member on the latest copy:
"...while the new one does seem influenced by yours, and may even have been trained on some outputs, it’s still a separate, newly trained LoRA. Training on image outputs isn’t something we currently restrict, and it's extremely common across the site.
Unless there’s a direct asset or dataset reuse, we can’t take any action on similarity of output".
I’d have appreciated more backup for original creators.
Maybe that’s on me for expecting it, as I’ve always respected other LoRA makers – if they drop something hot with a fresh idea and solid execution, good on them! I wouldn’t dream of biting their style or make a copy of their LoRA just because it’s successful.
But it seems that respect isn't a two-way street for some.
I’m not about to start naming names, but it’s clear some folks just don't get the concept of respecting the work you put into making a LoRA unique.
EDIT:
I have the urge to share this since some folks just think I mainly take images from the web and train purely on them.
So I share a tiny bit of my workflow here, so you may understand what goes into making a LoRA unique.
Left image is always my base, generated with AI tools like reve or other stuff.
Right image is the final training image after I was done with my inpainting and Photoshop session, that is then sent to training the LoRA.
Hope i could shed some light into that darkness with these examples.
So, here's the deal:
I'm taking my stuff off this site.
I'll be pulling down most of my LoRAs over the next days.
If you want to keep them, now’s your chance to download.
I’m keeping a few live tho, specifically for my good friends on here who rely on the Civitai generator because they can’t run stuff locally.
Will my LoRA's ever come back?
Never say never.
But right now, I’m a bit frustrated and sick of everything and I need to think about what I should do next, to upload them to my own site, or to other places.
I put a ton of hours into all these, especially crafting my own datasets, mostly from scratch.
Seeing that effort get blatantly ripped off just isn't it.
Hope you all can understand where I'm coming from.
In the meantime, I guess you can always play with the knock-offs.
If you made it this far down the article:
Thanks to all the people who used my LoRAs with respect and the way they were intended!
To help you express your art, to make killer images you love!
And also a big thanks to all of you who wrote positive comments on the LoRA pages, i read them all!
Truly thank you!
Much love as always!
ChronoKnight