Mostly NSFW, semi-realistic, flexible to use with LoRA.
This is my first mix, I used at least 6 of my favorite models starting with AbyssOrangeMix_hard, to draw out the semi-realistic style. Recommend using (worst quality, low quality:1.4) in negative, otherwise go crazy.
I tried, but with the first version you have a 50/50 chance to get a really good hand or the typical mangled fingers
I use primarily Clip Skip 1, ESRGAN_4x , sampling steps 20-30
Description
I'm pretty proud of this one, though It doesn't completely replace the previous. Through the use of the Merge Block Weighted extension this version adds in some texture and style variety. Primarily this comes from a Reverse Cosine weight merge with the Pastel-Mix stylized anime model. The style carried over well without losing much detail and flexibility from the previous version.
working title was weirdSperiment5.2
FAQ
Comments (8)
Where to get <lora:sy9i6v:1>?
How did you manage to achieve such a result on the 8th image?(https://civitai.com/gallery/74904?modelId=6260&modelVersionId=7951&infinite=false&returnUrl=%2Fmodels%2F6260%2Fcyriousmix)
I used all the same, but the result is different..
I think his negative prompts are embeddings, I can't really find those anywhere. There's 1 image in all his samples I can get close to and that's about it.
He seems to be using advanced prompting like a dictionary, and loras, hypernetworks, etc. You can get hints of that by analyzing his prompts
First of all, I thank you for your selfless sharing, and the effect of this model is very good. But I have a little suggestion to say. It is hoped that if it is a sharing model, please try to use the images generated by the model itself, and try to reduce the images generated by other doped things, such as the ' more effective ' works generated by doping lora. If you share such works, I hope you can also list lora together, otherwise the ai model with a very large gap between the pictures of the works shared and displayed will make the sharing of the basic model itself meaningless. Finally, thank you for sharing.
Thank you for sharing this model. I especially like your cover picture, but I have tried many parameters and can't generate similar colors and lights, as well as face. I found hairdSpercement5.2 and<hypernet: DIspersion: 0.563>in your parameters, but the name of the model is CyriousMix. Forgive my foolishness, can you share your Lora model? Thanks again!
Sorry that is a little confusing! weirdSperiment5.2 is CyriousMix1.4, I renamed it after I decided to release it. The Dispersion hypernet is https://civitai.com/models/1370/universityeuphorics-dispersion-hypernet
@Cyria thx XD