- This topic has 4 replies, 2 voices, and was last updated 6 months, 3 weeks ago by
.
-
Posts
-
Enhanced with Aiarty Image Enhancer, and edited in Photoshop.
Enjoy the artistic spree I am having these days 🙂
Lucky brother! I'd like a (step) sister like that.
While we're on the topic of AI, I went to train a dataset of 35 images yesterday, (100 repeats, 768px, 4 epochs), and it said 5.5 hours, bugger that. So I resized the images to 512px and the time went down to 1.5 hours. Turned out good regardless. Training a SDXL lora with 1024x1024 must take forever!@frank22 She's a bit shy 🙂
Strangely enough, I have never trained anything for AI 🙁
But, from what I heard, SDXL training stuff is a bit hellish unless you have a 3090/4090.
Thankfully, I found a model that seems to be able to do most of the stuff I am looking for, so I am using Loras (and the new Doras) very sparingly these days.
The trouble is that content for SD1.5 at Civitai is becoming much less and mostly Pony/XL diffusion base stuff. So I'm making more of my own loras.
@frank22 Yes, PonyXL based models are very good. I am using one in my latest batch of images, and it's perfect for what I like to make.
- You must be logged in to reply to this topic.
