Verified: 4 months ago
SafeTensor
Type | |
Stats | 60 0 |
Reviews | (11) |
Published | Dec 31, 2024 |
Base Model | |
Training | Steps: 2,200 Epochs: 10 |
Usage Tips | Clip Skip: 1 |
Trigger Words | dnksnst |
Hash | AutoV2 0BB4ECAB43 |
Trained on AnyLoRA noVAE.
This lora is very stubborn/overbaked (if using cfg1 otherwise its fine but not as I meant it to be) so any ADDITIONAL prompting must be weighted hard like 1.5 or so, and the lora-weight must be lowered to around 0.8.
CFG 1 WORKS FINE AND I DONT KNOW WHY?
It does what it needs to do on its own, mostly drawing icy landscapes with sunsets and the distinct shader/filter I found on MJ.
the bigger the image the easier it gets to fit in "foreign" objects.
trained on default 512x512
prompt with dynamic prompts extension:
<lora:DankSunset-10:{0.85|0.95|1.05|1.15|1.25|1.35|1.45|1.55|1.65|1.75|1.85}>, dnksnst,
<lora:DankSunset-10:{0.85|0.95|1.05|1.15|1.25|1.35|1.45|1.55|1.65|1.75|1.85}>, dnksnst, (forest:1.5), ice,