Nova Reality AM
Nova Reality AM is realistic Anima 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 Normal
Steps: 20~30
Clip Skip: 1-2
Denoising Strength: 0.65 - 0.8
CFG Scale: 4~6
Prompt: best quality, highres, (photo \(medium\):1.6), {Prompt}
Negative Prompts: 3d, worst quality, low quality, early, old, score_1, score_2, score_3, cartoon, anime, graphic, painting, crayon, graphite, sketch, abstract, glitch, text, signature, watermark, username, fused face, poorly drawn face, deformed, mutated, ugly, disfigured, bad anatomy, bad hands, missing fingers, extra digits, fewer digits, conjoined, long body, lowres, jpeg artifacts, simple background, bad ai-generated
Description
Initial Version
Anima-base v1.0 based
FAQ
Comments (5)
I tested v1.0. It is a very good semi-realistic model. The colors are beautiful to me, and the sheen I love so much really shines through. It is also excellent when it comes to prompt adherence; it does exactly what I write. Thank you for it :)
Title: Incredible precision, nationalities, and perfect depiction of details!
Review:
I tested the novaRealityAM_v10.safetensors model in an advanced way and I am extremely impressed by its quality and precision. I set the bar high: I combined detailed physical descriptions of characters with their specific nationalities, as well as real tourist data and attractions.
The result? The model handled it perfectly. Each character gained unique features appropriate to their origin, and the generated results beautifully captured every single, smallest detail. The model perfectly reproduces complex instructions, cares for background details, and excels at differentiating beauty types. The v1.0 release is a total bullseye. Highly recommended!
What's the recommended resolution to generate at with this model? 960x1680? That's an odd one... are there others, perhaps a 2:3 aspect ratio?
You can calculate these using 1280x1280 aspects
For 2:3 aspect:
ratio = 2/3
center_point = 1280
target_pixels = 1280**2
width = int(center_point * (ratio**0.5)) // 16 * 16 = 1040
height = int(target_pixels / width_ex) // 16 * 16 = 1568
So, 1040x1568 would do
@Crody Uh, thanks. But I was more looking for the resolutions the model was trained at. I tend to use pretty complex prompts and they turn out MUCH better when I stick to the numbers the model leans on heavily. Any deviation tends to get me lots of weirdness, even when I stay within the total pixel count boundary.