V1 Beta - Chud to Gigachad:
This is an early release, published because it already does the job, but it is not the final version. I think it can be improved further and plan to keep working on it. The main thing I'm not fully happy with yet: at strong positive weights the image can pick up a slight extra sharpness/clarity, most noticeable on skin texture. It's minor, but I'd rather say so than have you find it yourself.
Weight 0 leaves the picture as the base model draws it. Going negative softens and recedes the chin; going positive makes it more prominent and angular. It was trained specifically to hold up across different camera angles and distances , frontal close-ups, strict profile, three-quarter views, low- and high-angle shots, and full-body shots , rather than only a single tight head-and-shoulders framing.
Weight: -3 to +3
Description
Weight: -3 to +3
Known limitations
At strong positive weights, a slight extra sharpness can appear, mostly on skin texture. Still being worked on.
Every weight is a new drawing of the same seed, so small details unrelated to the chin (hair strands, exact framing) can shift slightly between weights, same as any LoRA slider.
FAQ
Comments (14)
Weird way to spell "Chud to Gigachad Slider", but I'll let you pass because I really do appreciate your work and overall creativity when it comes to what you release."
Honestly, I'll take the criticism، I'm not fully happy with this one myself. Already started on v2, which should come out cleaner. But hey, even as it is now it might work well for some use cases, you never know!
@shakir_abdo
I just want to clarify that I did not intend for my comment to come off as a critique, it's just that the examples you provided really do look like that and I find it amusing. I still have yet to test it on my end, but I am really happy to hear that you are looking to improve where you see possible. Your produce models that are really different in the sense that they focus on aspects of the face that really do help with eliminating the same-face syndrome I have been running into since SD 1.5. Krea 2 is great, and you only make it more great by uploading these very quality LoRAs. You are very absolutely amazing with this stuff.
@levinov_chyoa That means a lot, thank you! No worries at all, I didn't take it as harsh either. Really glad the sliders are helping with the same-face problem, that's exactly the itch I'm trying to scratch. Still iterating on this one behind the scenes to get it right before I call it done. Appreciate you taking the time to write this out!
Krea knows squidward, I wonder...
I'm not sure, to be honest.
@shakir_abdo Definitely doesn't seem to know handsome squidward, so that'd be a miracle.
Yeah, that went from no change at all to nightmare fuel. Express service train.
https://civitai.red/posts/31232607
Worked like a charm 😂✨
This is great for offsetting that weird AI-Girl face shape, however I see most of my generations have a cleft chin (butt chin). Maybe on the next version ,if there is one, apply a trigger word to smooth out the cleft, or make it more prominent. Thanks and great job!
For those encountering this, I am able to mostly offset it by adding something like "Their chin is pointy and completely smooth and oval shaped". Hope that helps!
Glad it’s helping! And thanks for the smooth oval tip , I’ll definitely experiment with a cleft chin trigger for the next version.
Sliders can be a bit tricky, especially when aiming for high quality results, since you’re balancing the feature change against identity and other image elements. So they often take a few tries and some tweaking to get just right!
@shakir_abdo Any luck with / chance for an overall head size slider? Maybe I'm just terrible at prompting (evidently...), but that might be helpful to keep proportions balanced.
That’s a really interesting idea! I’ll definitely give it a shot. Stay tuned ✨