LORA trained on Sister Lamia from MGQ.
Training based on ingame CG and fanart.
Trained on NED and tested also with AOMv2 and anything-v4.5 and more.
Would suggest weights around 0.5~0.9. (Above 0.7 it also pretty consistently reproduce the lamia form without need of other LORA, still using a 0.2~0.5 of anime-lamia can help with consistency)
Description
FAQ
Comments (3)
Could you do Eva or Morrigan next?!
Was thinking about Morrigan. I already made the shiki style lora to do shiki-bus like stuff but morrigan is more unique.
RIght now i am trying to understand the effect of heavy "by-hand" captioning when making LORAs to see if it is worth it and also working on new version for the other LORAs by including good AI images to have a more varied training set
If you have Discord, feel free to add me via the CivitAI group -- my friend and I have already set up some automation around data extraction for both the trilogies (though I guess Paradox isn't quite a trilogy yet...) and we had some TIs that still weren't quite right and were about to move onto trying out loras more when your uploads started appearing and were curious if you wanted to compare notes.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.