Description
Artist Style Lora, Purged off the site, I use it in my workflows and didn't want the community to lose it.
Prompting Info
Trained on Anything3 with Clip Skip 2, you can deviate from this but your results may get weird
If you want to really match the artist's style you should look for around 0.8 - 0.9 strength..
Buttt... frankly the images people are producing look way better around 0.65, it creates something distinctively different, with some elements of aka's shading style by following the original checkpoint much more closely.
Stick to soft shading models
Highres Fix is mandatory.. This LORA is slightly overtrained, it can hit the style really well but you need to clean up the jaggies with highres fix
Adetailer helps a lot too.. eyes can get weird without it
Negative Prompt
(worst quality, low quality:1.4)
Credits
Original Artist
Akabur - patreon.com/akaman & akabur.itch.io
Trainer
No Clue.. Everything Purged.
Description
FAQ
Comments (2)
If you are on the civit discord, there is a bot that posts whenever something is posted on Civit (along with the username). It's posts are useful for referencing things that were previously uploaded to the site.
Original LoRA model has been re-uploaded here : https://civitai.com/models/1339530?modelVersionId=1512689
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.





