Perpetual coffee request: Keep me motivated and caffeinated:
Description
FAQ
Comments (11)
Nice work. Thank you.
This looks great but I wish there was a SD 1.5 version.
I can't use XL loras on my setup but the sample images already look way better than the existing HR Giger/Biomechanical loras out there.
The world needs a quality Giger lora that's given the proper konyconi treatment
Glad you like it. Unfortunately, I've tried generating LoRA with this dataset for SD1.5 a while ago, and the results were bad.
Some sets with SD 1.5 won't work same as SDXL and viceversa. This for example could be the case. Kony will need different dataset and/or settings. However there are A LOT of GIGER's LORA's for 1.5. I did mine in the past but never shared it but there a lot of choices here.
@LDWorksDavid Yeah there are a lot of Giger loras for 1.5, just not a lot of good ones. None that can hold a candle to any of Kony's stuff
@Kodama didnt make he some for 1.5? I Remember yes...
@LDWorksDavid I did.. but I agree that this one is much better than my giger loras for SD1.5
@konyconi Totally, and since we are deep more and more on ComfyUI we are seeing SDXL to be superior in artistic terms to 1.5m anime not so much.
@LDWorksDavid Ah yeah there are some old ones that I didn't notice before. But also I kind of think SDXL is over-hyped and doesn't provide a ton of benefits compared to 1.5. Also not every UI supports XL models atm (currently using Easy Diffusion, which doesn't have XL support in its main branch)
@Kodama SDXL in art style generic/painterly surpass 1.5 EXCEPT in anime. In anime and flat concepts 1.5 still is better cause a more detailed VAE. Which IMO is the next step, to make a anime only VAE which cleans the lineas and flat/soften the image/imperfections. However I love 1.5 soften images too but if you ask me... SDXL is better with ComfyUI at certain datasets.
Another that doesn' put the info or checkpoint used on images example