Z-Image version is available!
Nova Reality
Nova Reality is realistic SDXL checkpoint with great feels of details
All images on example were created with diffusers with custom png tags
Rules
You cannot use the generated images for commercial use if it's not edited (or just turning it to black and white)
You can share images without any restriction if you don't monetize it
Advertising this model to outside is always welcome
Recommend Settings
Sampler: Euler a
Steps: 20~30
Clip Skip: 1-2
Denoising Strength: 0.65 - 0.8
(Pony)
CFG Scale: 5~7
Prompt: score_9, score_8_up, score_7_up, realistic, photo, detailed face, detailed skin, BREAK
Negative Prompts: score_6, score_5, score_4, score_6_up, score_5_up, score_4_up, source_anime, source_furry, source_pony, source_cartoon, blurry, jpeg artifacts, normal quality, worst quality, low quality, missing fingers, extra digits, fewer digits, bad eye
(Illustrious)
CFG Scale: 4-6
Prompt: masterpiece, best quality, amazing quality, very aesthetic, high resolution, ultra-detailed, absurdres, scenery, photorealistic, {Prompt}, BREAK, depth of field, photorealistic details
Negative Prompts: modern, old, oldest, cartoon, graphic, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured, long body, lowres, bad anatomy, bad hands, missing fingers, extra digits, fewer digits, very displeasing, (worst quality, bad quality:1.2), bad anatomy, sketch, jpeg artifacts, signature, watermark, username, simple background, conjoined, bad ai-generated
Description
Better posing and expressions
NoobAI EPS v1.1 + Illustrious v2.0 stable DARE applied
FAQ
Comments (6)
SO is ill or noobai?
Both and neither since it’s mixture of both with 0.5 ratio
DARE base was NoobAI
I'm really unsure what it is, with all of these realism merges, were it's horrible at retaining booru trained characters, and in the same exact ways.
Meanwhile your nova animal model stomps every one of these models in that department, but crumbles keeping the realism intact across complex prompts. This model for example is nice but, a step down in every way except realism from nova animal.
this is bizarre, and a reason i'm considering learning merging.
The primal reason why Nova Animal is greater in prompt adherence would be because of models I used for the merge
Nova Animal is mixture of several realistic models and Nova Furry thus the encoder is based on anime/furry
Nova Reality includes Nova Animal in the merge but to make the face more realistic, I couldn’t maintain its adherence
On future versions, I’ll try my best to maintain the balance
It will make sense to you once you start understanding how the data distribution affects the capabilities. It's a zero-sum challenge, you have a set amount of parameters and if you want to increase the density of one feature it inherently decreases the density of others. The hope is to generalize to the degree that it will understand every feature, but the reality is that it doesn't - at least not with current architectures and model size.
Listen prompt properly. Altrought i had some problem with "from side" perspective.