My first attempt to train multilora, so it's kinda unstable and a bit undertrained, but you still can get some nice results from it - a seva suits with different in-game factions color schemes
Known issues:
Lora is stiff regarding poses, don't be afraid to increase weight of pose tags if needed
You can get mixed color schemes, you can try to put some other color schemes tags in negative, increase weight of needed color scheme or just add something like "green theme, green color scheme"
Tags "seva_loner, seva_eco_green" are almost not working, "seva_merc, seva_freedom" are very unstable, "seva_mono" isn't game accurate, but still recognisable and works fine, "seva_duty, seva_clearsky, seva_eco_orange" are most stable and best working tags
Recommended weight is 0.85-1.0
Description
FAQ
Comments (3)
Did you tag the game screenshots "3D"? The suit seems odd against more complex bgs. Also recomend using arbitrary unique triggers: I think it's reading "suit" in the trigger and making weird hybrids.
Thanks for the feedback, the model 100% needs re-training, and I'll use these advices for sure<3
@GlavaKolhoza Yeah I've had pretty good luck with models trained on 3D models, especially for characters with more realistic proportions, but I can generally immediately spot when someone makes a model with them without tagging the pics 3D.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.