PhotoPedia XL

photopediaXL_45_full_fp32.safetensors

photopediaXL_45_pruned_fp16.safetensors

Photopedía xl

Type 4

Type 3

Type 2

Type 1

<ul><li>4.5 is out. Very happy how it turned out!</li><li>MBW'd the living hell out of this one. Must have went back and forth through more than 30 merges to get here.<ul><li>You can right click the images and open them in new tab and zoom in quite a lot to check smaller details.</li></ul></li></ul><ul><li>Addressing certain issues such as hands / eyes. Eyes are much better. Hands are better as well. Generations can still produce some nasty monstrosities though</li><li>Slight overall feel change</li></ul>

GENERATION PARAMETERS & SUGGESTIONS ( As of 4.5 )
ℹ️ 4.5 generations were made with the following settings

H/W: 1024/1024 - 1024/576 - 576/1024 - I tend to stick with 1024/1024
Sampler - DPM++ 2S a Karras is the main sampler I used for 4.5
Steps - 40 - Can go lower
CFG - 4-9 - I went lower than usual, usually 4-5

Generate an image with Hires enabled

Hires upscale: 1.5 
Hires steps: 15-20
Hires denoising: ~ 0.6
Upscaler: 4xLSDIRplusC

Send to Extras. Upscale once by 1.5. Send the upscaled image to Extras again, 1.5 upscale again. I usually run this in batches.

Upscale 1: 4x Foolhardy Remacri - 1.5 upscale in Extras tab
Upscale 2: 4x NKMD Superscale

I have not used ADetailer, negative embeddings or loras.
ℹ️ MAIN PROMPT

photoshoot, photo studio, RAW photo, editorial photograph, film stock photograph, cinematic, posing, amateur photo, analog, raw, f2, 35mm, an (amateur photo), flash photography, taken on an old camera, polaroid, 8k, highly detailed, (high quality, best quality:1.3), Extremely high-resolution, film grain, Kiki from Netherlands, long (waist-length:1.2) combed auburn hair, ((macro lens)), dark atmosphere, ((focused on eyes)), feminine expressions, photography, dslr, 35mm, Fujifilm Superia Premium 400, Nikon D850 film stock photograph, Kodak Portra 400 f1.6 lens, 8k, UHD

Try to run a generation without any negatives first. I am personally too lazy to test each one of them and their impact on generations and the current prompt did just fine x)

Try to run a generation with very simple prompts before you start adding more

ℹ️ I did not use any negative embeddings or loras to change the outputs. Should you use them? As I always say, feel free to experiment, I am pretty sure that some of the negative embeddings will work very well with the model. I generally try not to use too any.

ℹ️ INPAINTING



Files with SHA256: 3865942aef6823419dc69c82c44eda49be1f15986cb27dafc61c9b6827750298