A SDXL 1.0 LoRa made by AI_Characters for improving the photographic capabilities of the base SDXL 1.0 model. It makes the photos look more realistic, less fake and AI generated, and overall just higher quality - with slight improvements to NSFW as well.
Usage recommendations
Weight: 1.0 (may need to be adjusted down when using in conjunction with other LoRa's) Model: SDXL 1.0 with 0.9 vae Generation parameters: 35 steps | DPM++ 2M Karras | CFG 7 Negative prompt: text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, distorted, blurry, glossy, blur, jpeg artifacts, compression artifacts, poorly drawn, low-resolution, bad, distortion, twisted, excessive, exaggerated pose, exaggerated limbs, grainy, symmetrical, duplicate, error, pattern, beginner, pixelated, fake, hyper, glitch, overexposed, high-contrast, bad-contrast with only the "G" text encoder enabled (see my article here for how to do that)
Trigger words
"photo in phst artstyle"
"with beautiful lighting"
Changelog
v2.0
Complete revision - trained with a much lower learning rate but more epochs, leading to increased likeness, flexibility, and variety;
fixed the eyes issuethat may have been a preemptive victory celebration - eyes are definitely better now but still a bit worse than base SDXL; slightly improved NSFW
v1.0
Initial release
Additional Info
Credit is not mandatory but I would still appreciate credit when posting images using these LoRa's on social media! Feel free to @ my Twitter and/or Instagram accounts too (@ai_characters)! I would love to know when it is used!
I share all of my model's and training knowledge for free, but donations to my Ko-Fi would be greatly appreciated! Testing so much and training so many model's costs a lot of money (and time).
Feel free to also try out my RVC2 AI voice models which you can find here!
See here for an extensive article about my entire training workflow and configuration if you want to train your own SDXL 1.0 LoRa's and achieve results similar as me!
Stats (v2.0)
Dataset:
311x photos
Epoch chosen for the final model:
35
Learning rate:
6e-6
For more info - including the training configuration used - see the article linked at the end of the "Additional Info" section.