UPDATE: Re-trained on SD 1.5 - It doesn't retain the comic book style UNLESS you use it on an illustration/comic model. YOU CAN NOW EASIER do realistic. The HEADGEAR works better now, it's not perfect. Consider this still BETA because y'never know we may go finding more data.
Trained on NAI, after also training one set on MALWARE DREAM realizing you just don't train it on super customized models. (Yes, you can ask to test the other version, it's just not up on civit.)
Right now this is considered a "BETA" - as in you are going to see things that any OCD comic book fan would throw a book at us for:
Consistently does her headgear incorrectly, mistake in the data (Yeaaaa when you find some really cool data and realize it's edited movie pics of Wanda not Lorna)
Struggles between outfits because we didn't have consistent data of ANY of her outfits, due to not wanting to spend forever and a day training it (Need to ask more about LORA TRAINING) - With this, experiment how you will on "BODYSUIT" and "LEOTARD" - because they both do different things
Also clothing issue: HER MORE CURRENT "BODYSUIT" outfit is VERY close to her old 90s one. She tends to get EASILY mixed up with Enchantress AND VIPER both - as well as other green haired chicks.
Body types WILL VARY even on the NAI version (which is this one) - We had a wide variety of data, including BBW/Voluptuous cosplayers - and welp, SD BE SD!
Hair TYPES varying from decade to decade, we've tried to limit EYE COLOR and HAIR COLOR issues by absorbing those into the activation tags.
How we're going to rectify this: Do more data study, filter the data. It's clearly not ENTIRELY a tagging issue, though I think on clothing that MIGHT be.
Note: You're at risk for any legal implications from USING this model, be aware of all copyright laws in your country/state whatever, and also do not do EXPLICITLY ILLEGAL content with this LORA.
If you use it commercially, be aware that you may be at risk for "COPYRIGHT"/DMCA.
Generation sites: Just let us know ahead of time before you yoink it, we're gucci about this mostly now - we want to relax, chill and then link back to the places that think our stuff's worth it. We don't do exclusivity, so just yea let us know!
If you got requests, or concerns, We're still looking for beta testers: JOIN THE DISCORD AND DEMAND THINGS OF US: https://discord.gg/5t2kYxt7An
Lora Request Form: https://forms.gle/NEkcMmVEjG4pBAP7A
JOIN OUR SUBREDDIT: https://www.reddit.com/r/earthndusk/
Listen to the music that we've made that goes with our art:
https://open.spotify.com/playlist/00R8x00YktB4u541imdSSf?si=b60d209385a74b38
We stream a lot of our testing on twitch: https://www.twitch.tv/duskfallcrew
Pre-Release Models, Lora Backups - Send us pizza! https://ko-fi.com/DUSKFALLcrew
Submit SFW and amazing wallpapers at PaprGG: https://discord.gg/2UXkGwndVE
LOOKING FOR THE M4DREAM IN CASE IT ENDED UP IN THE SAMPLES?
https://huggingface.co/Duskfallcrew/lora_dump/tree/main
Head to the link above and look for "POLARIS-M4DREAM.SAFETENSORS" - it's the one trained on Malware Dream, and it DOESNT BEHAVE as nice as the NAI one in this repo.