According to my latest findings, not recommended to download.

LoRAs seem to have a really strong influence on the checkpoint, unlike a model/model merge...

but I already have a plan! it just takes a while... to be continued >

Version 1: The Experimentation Chronicles

Confession Time:
I must admit, I’m navigating this process with a healthy dose of uncertainty.

The Mix-Up:
I merged this checkpoint (my first ckeckpoint from huggin face, sorry no idea which one it is anymore) with my LoRAs and here’s the outcome!

The Craftsmanship:
The sample images were crafted using ComfyUI. Using double KSampler at 816x1232, upscale latent by 1.5 to reach 1224x1848. KSampler 1 denoise 1.00/KSampler 2 denoise 0.50.

Important:
Sampler: dpmpp_2m_sde | CFG: 3.0-4.0 | more Step´s more Details (dpmpp_3m_sde seems to have an urge to create living beings)

Next Steps:
I’m excited to dive into Civitai next and see how the results stack up. Stay tuned for more delightful surprises!

Visions:
Merge with new high-quality LoRAs that I will create. Find the right weighting for optimal results

Negativ Prompts:
spheres (can help with too many round elements) probably too high weighting of shattered void LoRA

My Reminder:
set down weighting of "Shattered Void" LoRA maybe to 0.4 or 0.3.

set down weighting of "Dark Beauty" LoRA or train new with better eyes.

try to include a VAE AT FIRST!.(No idea if it’s included in the checkpoint).

“nice” LoRA training, too much focus on dark fantasy.