Version 2.0 has been trained a little differently and with more iterations in a attempt to improve the overall art style.
I did attempt to do a ZImage LoRa for this model, but the training didn't work out.
Right now all of my training is handled exclusively through Civitai's cloud tools. Which is nice in some ways, but it comes with built-in limits on iteration speed, dataset experimentation, and model flexibility.
My goal is to shift to a fully local training workflow so I can deliver deeper, more refined LoRAs and start training full checkpoints without cloud bottlenecks. To make that happen, I’m working toward upgrading my rig with an NVIDIA RTX A4500.
If you enjoy my work and want to help speed up this hardware milestone:
⚡ Tip Buzz directly on this post or my models
☕ Support on https://ko-fi.com/diffuseddamsel
⭐ Review & share your gens—engagement helps boost visibility more than you know!
Thank you all for the continued support!
