Update v3.0 on : https://civarchive.com/models/2358786/zib-daily

There is two new LoRA, but i recommend to use them together on ZimageTurbo or Base:

• Zimage-dAIly - 0.4 Strength - Trained on Zimage Base, more close to my OG LoRA but with the inconvenient too, but in a low value he gives the little contrast boost i like without impact negatively the image.

• Z-image-dAIly-2602 - 0.6 Strength - Trained on Zimage Base, "lighter" than the one above but more permissive, less default, cleaner.

As they are trained on Zimage Base, you need a "trigger words" just to cancel the photorealistic reflex of Z image, "Anime style" "Graphic anime illustration" "2D Drawing" "3D model Rendering" it doesn't matter as long you send one to push Zimage out of his realistic habits.

Please do not create new LoRa/models with the generated images without asking :D

V2 : NO TRIGGER WORDS ANYMORE.

No trigger words, I advice to use gauss legendre 2s + bong tangent as sampling. The sweet spot for speed/good details on my 4080 16GB VRAM is 2.2 Mpx for initial generation, but far away toes make it struggle :/

v1 : I use it at 0.85-1 strength, trigger word : "dAIly artist". But on ZIT you should use a full sentences if you got a detailed prompt. I usually insert "(The graphic style of the image is dAIly artist:1.3)" at the beginning and "According to dAIly artist graphic style, the stunning young woman is the only character on the image." at the end. You can trigger the style easily without this. It's just more ... efficient.

So, this LoRA is a result of several training passes, on many models, from SDXL to FLUX then WAN22 to ZIT, and now... on the Z-Image Base.

It's main purpose is to generate waifu pictures