General Illustration Model
Wow, I had this laying around huggingface without even uploading it.
Even funnier still? I HAD AN FP16 Safetensors already in my drive, and i've REALISTICALLY CHECKED AND RECHECKED if this was a double upload and guess what NO!
IF YOU NEED A PRUNED VERSION OF THIS HOLD FIRE, I'm running on colab free, and i have a few credits on Runpod left, but i need to chunk out a payment on my psych bill, and not my AI funds - unless you're happy to throw a few coffees at us LOL <3
THIS WAS NOT TRAINED ON GUYS I DONT THINK? So "FACIAL HAIR" in the prompting was a lagover from the merge that DOES have males in it that isn't this model, so ignore that in the image prompts.
TEST THE MODEL YOU SAY? you can test it here: https://huggingface.co/spaces/Duskfallcrew/duskfall-s-general-digital-art-model
*(it's running on CPU, but if you want it go on GPU hours maybe like sign up to ko-fi or patreon?)
We stream a lot of our testing on twitch: https://www.twitch.tv/duskfallcrew
any chance you can spare a coffee or three? https://ko-fi.com/DUSKFALLcrew
If you want custom LoRA OR MODEL trained an option will become available on the Patreon: https://www.patreon.com/earthndusk
Lora Users:
I've removed the Lora from the main repo and put it here: https://civarchive.com/models/18429/general-illustration-lora-lycoris-extraction-by-kappa-neuro
Earlier the LoRa you were seeing was here: https://civarchive.com/models/14846/duskfallcrews-art-style-lora
UPDATE: @Kappa_Neuro has graciously extracted LoRA from this and i'm adding it to the pile of downloads Please note that a FULL model safetensors is still coming the safe tensors you see is the EXTRACTION. Also due to the fact it is LYCORIS extractions you'll need the plugin here: https://github.com/KohakuBlueleaf/a1111-sd-webui-locon
SAFETENSORS AND POSSIBLY PRUNING to come - please let me know if theres' anything wrong with me XD
NOTE! This model is a little unhinged, but with the right care, babysitting and psychotherapy, IT WILL RESPOND to Waifu and Danbooru tagging.
Don't ask me how, it was an old HF trained one, but i'm testing it now and it's smooth as a milkshake from an UNBROKEN DAIRY QUEEN BLIZZARD MACHINE!
SAMPLE PROMPT
(plus embeds, negs and hi res fix deets)
gendigi, solo, (masterpiece, best quality, ultra-detailed, highres), perfect face, sidelighting, lustrous skin,(bloom), (shine), lighting, ray tracing, sci-fi, (red hair), multicolored hair, portrait, blue eyes, black hair, undercut, jacket, beard, looking at viewer, smile, streaked hair, highly detailed, depth_of_field, very detailed background, highly detailed background, Masterpiece, Ultra detailed, great composition,Dynamic angle,[Bottle bottom],(wide shot), extremely delicate and beautiful,(Highest picture quality), (Master's work), depth of field, solo, extreme light and shadow, masterpiece, rich in detail, (fine features), (highest quality), (masterpiece), (detailed eyes), (beautiful) detailed, beautiful detailed eyes,(straight-on), full body, cyberpunk:1, lightning, (extremely detailed CG unity 8k wallpaper),(masterpiece), (best quality), (ultra-detailed), (best illustration),(best shadow),perfect lighting , perfect anatomy , vivid colors,gendigi
NEGATIVE PLUS NEGATIVE EMBEDS:
verybadimagenegative_v1.1-6400 bad_quality bad-hands-5 bad-image-v2-39000 vile_prompt3 By bad artist -neg, erotic, nsfw, anatomy, bad hands, text, missing fingers, extra digit, fewer digits, signature, watermark, username, blurry, artist name, abstract, picasso, amputee, deformed, blurry, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, extra leg, bar censoring, blurry face, out of frame, border bad-artist-anime bad-artist
MUSIC:
I'm not gonna promote my own damn music on here LMAO - go look up "DUSKFALL" on youtube, and NOT THE METAL BAND, if you see stuff about "STABLE DIFFUSION" and Second Life in the same set of videos that's us.
VAE:
At the moment i'm using the one that is with Kenshi's and that's the one i'm going to upload - but because VAE wildly effects how your outputs happen? I'm doing you a favor: USE WHAT YOU WANT! It's all up to you.
DO I NEED HI RES FIX:?
No- but it's entirely up to you:
Examples of NOT HI RES FIX:
HI RES INFO:
Description
It wasn't trained on my art, so the prompt was ALMOST useless because there IS a merge with our art in it that i'll release after testing.
FAQ
Comments (24)
Is this a lora? Why is it 144mb?
GOOD QUESTION did GOOGLE DRIVE LITERALLY EAT MY MODEL or did i upload the wrong thing? Temporarily until i get this fixed there's an HF Link to the OG file: https://huggingface.co/MultiversexPeeps/duskfall-s-general-digital-art-model
FIXED! Accidentally uploaded a LoRA instead of a full model lol, it was in the same row in my downloads folder lolol
Can you upload a safetensor?
Defo, i'm a bit behind on those but in the coming days i'll have one up for this for you!! Thank you for asking <3
Can you add, if possible, the 2gb pruned fp16 no-vae safetensor version ?
I'd also like the LORA version; can you reupload it again ?
are you asking for a BAKED vae or a no VAE? i didnt train this with a VAE, this was using huggingface, i'm not sure what you're asking for that - and i'll link you the Lora
I THINK IT WAS THIS: https://civitai.com/models/14846/duskfallcrews-art-style-lora if it wasn't this then .. OH HOLD ONJ LET ME CHECK MY DOWNLOADS FOLDER LMAO It might be the one someone extracted
Yea it's the one I linked, XD i accidentally re uploaded our own art LoRA
Safetensor is already planned btw, but i dont get what you mean by PRUNED lmao - do you mean "PRUNE IT SO YOU CAN TRAIN ON IT" (you know i'm math and SD stupid by now XD)
@duskfallcrew About the pruned fp16, if can be of any help (not intended for training):
- https://github.com/Akegarasu/sd-webui-model-converter
"convert to precisions: fp32, fp16, bf16 - pruning model: no-ema, ema-only - checkpoint ext convert: ckpt, safetensors"
- https://github.com/arenatemp/stable-diffusion-webui-model-toolkit
"Cleaning/pruning models - Converting to/from safetensors - Extracting/replacing model components"
Another user, in another model page, said:
If you're using automatic1111 webui, you can do the following:
Go into checkpoint merger
Input this model as Model A (nothing else)
Name it (Something like "ModelMerge_Pruned_fp16")
Select "No interpolation option"
Select "safetensors" checkbox
tick the "save as float16" checkbox
Merge
What you'll get is a .safetensors file which is more secure than ckpt, 2GB~ in size and the generation is exactly the same as the .ckpt file.
@duskfallcrew about the baked-vae models I prefer to not have it, for better flexibility.
About the LORA I was referring to this model, since id adds a better way to mixing functionalities of different models, if compared to the traditional merging way that always produces huge 2-4-7 gb blends.
@ritcher1 There's no LoRa version of this dude, LOL it was actually https://civitai.com/models/14846/duskfallcrews-art-style-lora uploaded to this model file. XD the pictures you see are entirely up to the prompting in the generations. The file should be Fp16 ANYWAYS when it comes to safetensors, ebcause it's a huggingface trained one, tehy don't do fp32? XD i'm gonna ask on this though, i don't use UI to convert, i use a adhoc verison of it on HF that cameronduru made.
@ritcher1 So basically - you don't need a pRUNED file, you need me to convert it - If you'd check the size because this is a trained at HF file, not a notebook one, it's a 1.99 gb file - it doenst' get smaller than that. If you need a LoRA off this, i'm sorry i do'nt even remember what the training data is, and i don't have the ability to extract LoRa right now
@duskfallcrew Ok, no problem.
Update: I'm dumb. Evidently you SHOULD AND CAN prune a 2gb model.
Ask me to microwave it for 2 minutes, leave t out in the sun, feed it after midnight andi'll see if i can't nobake a cheesecake before midweek fro you XD
@duskfallcrew As I said from the start: "Can you add, if possible". If not, no problem. No pressure at all.
@ritcher1 Yea i'm sorry, LOL, NoBake Cheesecake coming when I stop being so ADHD. Not that i'm trying to use excuses for being cranky, but if you DO have requests for stuff .... HMM wondering where there might be a better place to put Model requests besides like... bugging people to do ko-fi XD
Ok, i've got the FP16 EMA safetensor, this one doest' have a VAE baked in, i don't tend to do that cause i love mixing my VAE - TURNS OUT I HAD A SAFETENSORS IN MY DRIVE FOR OVER A MONTH. Query: Next time i have a model like this that i forget to upload, can you bap me with a frying pan and ask me if the safetensors is in my backup drive? LOL
Your LORA version is tagged as Model instead of LORA just FYI
I'm aware of this LOL, i didn't upload it seperatley xD i didn't test it yet.. i'm dumb :P
In saying that the LoRA versions are the previous version and are extractions and are half or more of the size of the OG model lol
@duskfallcrew Yeah just wanted to let you know because it'll mess up people using the API to snag models.
@entmike2 AAHH CRAP thank you!!! I'll snap them off the branch and just share them seperate!