Merged with Hyper dev 8 steps lora from Bytedance, Unet only, need clip L, t5xxxl and vae
https://huggingface.co/ByteDance/Hyper-SD/tree/main
Ultimate version
More versatility
Improved artistic side
V4.0
Better realism
Better versatility
Better nudity
Try it with:
Euler/Beta, 10 steps, guidance 3,5 and detail deamon
V3.0
More lora merged for more versatility
Better lighting
Fix vertical trame
V2.7
Merged with my lora: https://civarchive.com/models/866492/moreface-lora
More realism
More versatility
V2.6
More realism, thanks to @bissonfrederic69429 for his wonderful lora SkinDetails
Recommended settings : 10 steps, Guidance 3, Euler/Beta
V2.4
More lora merged
Improve NSFW
More realism
More versatility
Hyper Dev gguf-q4_0 version merged with my lora Real-lora.
Hyper Dev Fp8 Unet version merged with my lora Real-lora.
Hyper Dev bnb nf4 Unet version merged with my lora Real-lora.
https://civarchive.com/models/650205?modelVersionId=727461
More realism for skin, background with less blur.
8 to 20 steps recommended settings : 12 steps, distilled CFG 3, Euler/Beta
Node for comfyui: https://github.com/DenkingOfficial/ComfyUI_UNet_bitsandbytes_NF4
Work with Forge, add vae and clips manualy.
Description
gguf version
FAQ
Comments (13)
What is the difference between the guff q4 and unet versions? For some reason, all flux models work at the same speed on my 1660 (6 GB) - 1 iteration per minute. I first connected q3 model which should fit in 6 GB, then your q4, now the unet version - the speed is the same, only the quality is better. Strange. The video card processor is fully loaded in all cases. It seemed to me that the model from the video card memory should work faster (q3 ~ 5gb), but this does not happen.
This is not normal, the fp8 version which is 11gb, cannot be loaded completely in vram if you only have 6gb, it must therefore be divided with a part in vram and another in ram, there is then a swap is carried out which necessarily slows down the total rendering(not iteration per minute)
@jice this is also the same for me on a rx 6600 with 8 gb of ram and , I get the same speeds for 11gb models vs all the gguf models I tried. So why use the gguf then ...
@xpnrt Sorry for you
hyper dev gguf .. howmany steps needed ?
10 steps, the same for the fp8 version
If you can do it in 10 steps with Euler, you can probably do it in 8 steps with another scheduler. It may or may not be worth your time experimenting to figure out which one(s).
@lesjo yea thanks
oops. it's produces black screens in forge.
@ckotty3249 Change your vae, i use forge only without issues
Hi! how to generate old person? I set the age in prompt, but all the honesty is giving out young people.
@Dozor "woman with 60 years old" work for me without issue, look at the gallery
Yeah this model sort of exhibits "everyone is young or old, nobody in between", but that seems to be how all models are biased to a degree. This one does it more than most though.