50 styles trained in one lora, more of an experiment using mostly requested artists and some that I already had a dataset for. Most of them are nsfw artists and there's one or two that don't really look like the artists, probably from my dataset. Artists list is in the training data file.
Description
FAQ
Comments (19)
Can you list the artists?
I downloaded the training data file and it has this list (49 artists):
by terasumc
by butcha-u
by majalis
by dinoartforame
by mossacannibalis
by omone hokoma
by rkrk
by ebonyxh
by yaegashi nan
by ohisashiburi
by tsubasa tsubasa
by obui
by hershuar
by mogudan
by kase daiki
by popogori
by poper
by oeakakizuki
by raita
by reoen
by alkemanubis
by jlullaby
by scottie
by tekuho
by s16xue
by shuz
by rustle
by opposum
by greatmosu
by forastero
by ningenekaki
by asanagi
by as109
by nyantcha
by muk
by mimonel
by laserflip
by kabuyama kagi
by artistgloss
by gemba
by fizrot
by fkey
by caisan
by asakuraf
by aowvn
by andava
by aki99
by kapibara
by tsuki suigetsu
@bionagato nice artists, alot of goat's!
Not sure what's wrong with it, but it only shows up in my list of available loras when I select "show incompatible loras" option in automatic1111. Check the embedded data and it returns errors and the outputs are messed up.
Try reinstalling the lora, works fine on my end.
@Ligmanese Still not working, its acting like its not being recognized by webui but this is the only lora I've ever had this issue with lol. Are you using automatic1111?
@sliptrick I'm having this same issue.
same issue here
I tried to add the artist tags only to the metadata for easier use, so I guess it failed and caused issues with the webui lora browser. I'll fix it if I ever do a v3. The lora itself should work fine though, just copy the name.
So I guess I figured out my problem, adding any amount of prompt weighting makes the image output melt lol @tjohnson424 @Aruhshura might be the same problem?
@sliptrick that's an issue with the webui and SDXL, try this extension https://github.com/hako-mikan/sd-webui-prevent-artifact
@Autismix_anon Don't need that in forge any longer. Just go to settings and select Emphasis mode -> No norm. Might even work better, since I was still getting glitches with extension, but haven't had any since this.
@Autismix_anon @sliptrick I managed to create a (very makeshift) fix for this by editing the embedded metadata of the LoRA file. The issue seems to be that the LoRA file is missing almost all metadata, which causes the automatic1111 webui to error when trying to display the metadata.
For anyone interested, here's how to fix it:
1. Get a LoRA metadata editor, I used https://civitai.com/models/249721/lora-metadata-viewereditor
2. Load the LoRA in the editor
3. Replace the metadata with the content at the end of my comment (In the "Metadata Editor" tab)
4. Update the LoRA ("Update & Download" button in the LoRA browser editor)
5. Use the updated LoRA file as you normally would
Fixed metadata (copy everything in between the curly brackets, including the brackets themselves):
{
"ss_output_name": "hyperautism_lora_XL_dim32_8e_v2_civit",
"ss_base_model_version": "sdxl_base_v1-0",
"ss_sd_model_name": "ponyDiffusionV6XL_v6.safetensors",
"sshs_model_hash": "6090fe72b506da41b6595518afa5911c4817e883a4cd5de26af55fe53df8dc23"
}
@Autismix_anon I have that already lol, I can use any other lora with this prompt but this one, for some reason, melts outputs if I have anything weighted above 1.1.
Its not an issue, personally, this is an experimental lora, after all. I was just letting people know that could be their problem.
Just in case anyone still around here, changing the "SD Version" in the JSON for this LORA (within the LORA folder in A1111) to "SDXL" made it appear for me.
If not, someone commented (sorry, couldn't find the comment) about the option to enable "incompatible" LORAs from the settings. Makes it appear as well and you can manually change it to SDXL from the GUI (in case you DO have incompatible LORAs?).
Anyway, hope it helps cause when StevesAISlop mentioned metadata editors and steps, my face became a Picasso painting XD.
It didn't show up in the avilable loras for SDXL while I using forge UI. Why does it happen?
What happened to some of the artists from the attached image? Fishine does not seem to be in the model.
The List
terasumc, butcha-u, majalis, dinoartforame, mossacannibalis, omone hokoma, rkrk, ebonyxh, yaegashi nan, ohisashiburi, tsubasa tsubasa, obui, hershuar, mogudan, kase daiki, popogori, poper, oeakakizuki, raita, reoen, alkemanubis, jlullaby, scottie, tekuho, s16xue, shuz, rustle, opposum, greatmosu, forastero, ningenekaki, asanagi, as109, nyantcha, muk, mimonel, laserflip, kabuyama kagi, artistgloss, gemba, fizrot, fkey, caisan, asakuraf, aowvn, andava, aki99, kapibara, tsuki suigetsu
I am unsuccessfully trying to make a multi style lora. Can you tell me the settings you used? Can't see the metadata from yours
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.