Early access to DesireX3 is available on my Telegram.
The Amateur version is also available by subscription to my private channel.
Designed to look innocent, but she has attitude.
Every image starts with natural light, realistic skin, and a photographic feel.
She already has a personality, so treat her nicely.
Ignore every suspicious LoRA you found at 3 a.m.
Ready for an easy first date? Grab the wolFkroW_SSB.
Expect only the essentials: a few nodes, a prompt, and you’re on your own.
X marks the spot - if things get weird, tell me what you did. I know her.
Links:
Buy models or support - https://t.me/blacksnowskill
Description
FAQ
Comments (6)
Loved your previous versions. Unfortunately, Amateur_INT8_ConvRot might have had its amateur lora baked in a bit too strongly. It saps all color from an image, and more often than not the final image is blurrier than expected.
It’s a matter of taste, after all. Some people need exactly this kind of behavior from the model out of the box. I posted it as an experiment to see how much the community needs it and whether it’s worth developing this branch of the model.
@clueless_engineer You can merge it 50/50 with the official int8, and it will be half the strength. Or merge it with another nsfw checkpoint. I have made myself a 50/50 moodyV8/Desire merge that makes some incredible results imo.
@Loraholic Quite resourceful. I also recommend mixing it with the original DesireX model to ensure maximum compatibility.
@Loraholic can you merge int8 with int8 krea2 models?
@doombatz you can merge every model. I have merged bf16 with int8 etc.
Just need to convert them before merge, and back to int8 again.
Use Opus to make a simple script. If the int8 is not a convrot-model there will be around 7% error in the worst layers compared to bf16. But when the int8 model uses convrot its only around 2% dataloss in my testing.