license: creativeml-openrail-m
tags:
- Text-to-Image
- stable-diffusion
- lora
language:
- en
- zh
Stable Diffusion NAI lora Index
中文文档 : 还没写
lora is one efficient way to fine-tune a stable-diffusion model.
It can result in a model smaller than a dreambooth model. And provide better result than a textual inversion outcome(embedding).
This repo is for indexing NovelAI related LoRA works in huggingface.
Preview the "good models" with no "explaination" in their repos using a "benchmark".
Note that NSFW will not be included since I cannot put previews.
And there are plenty of lora models, I may only include "good models" based on my judgement.
Please contact me or create pull request if you find something should be included.
Click "Like" if you think the index is helpful~
dranzerstar/SD-textual-inversion-embeddings-repo
Link: https://huggingface.co/dranzerstar/SD-textual-inversion-embeddings-repo
models: char-416-space, char-antonia-og, char-april, char-cms-gn, char-cms-og, char-florine, char-g41space, char-gronru, char-kyaru-gn, char-madoka, char-qu, char-siobhan, char-sop2anni, char-sophia, char-sp9-og, char-toka-dancer, char-toru-generalized, char-toru, char-vepley, char-zas-wq, char-m16a1 (I cannot generate)
ikuseiso/Personal_Lora_collections
Link: https://huggingface.co/ikuseiso/Personal_Lora_collections
Note that the owner claimed that the weight should be set to 0.6~0.8.
So, I take 0.7 for all the previews below.
models: ame-chan_needy_girl_overdose, chouzetsusaikawa_tenshi-chan, grea_shingeki_no_bahamut, iono_pokemon, kisara_engage_kiss, laundry_dragonmaid, lishenna_omen_of_destruction, lovely_labrynth_of_the_silver_castle, lucy_cyberpunk, Miorine_Rembran, ralmia_sonic_racer, seulbi_lee, suletta_mercury, vampy
sylveriate/lora-characters
Link: https://huggingface.co/sylveriate/lora-characters
Actually I cannot generate good results from this repo, maybe the owner can teach me how to do it...
models: abigail, emilico, hiroikikuri, kateshadow, pippa, selen, shondo-improved, shondo, shondo-improved, tenma
backslashlim/LoRA_Dumpster
Link: https://huggingface.co/backslashlim/LoRA_Dumpster
Link Only
These owners are managing their model cards quite well.
So, I will only copy the Links and summarize keywords.
Link: https://huggingface.co/YoungMasterFromSect/Trauter_LoRAs
Genshin Impact : Eula, Barbara, Diluc, Mona, Rosaria, Yae Miko, Raiden Shogun, Kujou Sara, Shenhe, Yelan, Jean, Lisa, Zhongli, Yoimiya
Blue Archive : Rikuhachima Aru, Ichinose Asuna
Fate Grand Order : Minamoto-no-Raikou
Misc. Characters : Aponia, Reisalin Stout
Artstyles : Pozer
Link: https://huggingface.co/breakcore2/loras
Amber, artoria pendragon
Link: https://huggingface.co/Onusai/LoRAs
Chainsaw Man
Link: https://huggingface.co/Aotsuyu/ozen-Lora
Ozen Sama
Link: https://huggingface.co/Doctor-Shotgun/ds-LoRA
Persona5: Ann Takamaki, Futaba Sakura, Haru Okumura, Kasumi Yoshizawa, Makoto Niijima
DONOT KNOW THE SOURCE
You can download the models in this repo, if you know the source, please contact me. I will change them to links.
qiqi
Trigger token is "qiqi", also "1girl", "solo" are recommended.
# Sample prompts
1girl, qiqi, jiangshi, qing guanmao, coin hair ornament, ofuda, bead necklace, qiqi \(genshin impact\)
klee
"1girl, klee"
"1girl, nsfw, klee, solo, completely nude, hat"
"1girl, klee, solo, ahoge, hat feather, twintails, gloves, boots, bag, red dress, bangs, backpack, dress, brown gloves, knee boots, low twintails, brown footwear, cabbie hat, red eyes, pointy ears, long sleeves, clover print, scarf, socks, hat, blonde hair, kneehighs, coat, charm \(object\), brown scarf, hair between eyes, red headwear, light brown hair, bag charm, randoseru, sidelocks, hat ornament, bloomers, pocket, red coat, white socks, feathers, white feathers, klee \(genshin impact\)"
nahida
nahida_cross-shaped_pupils_finger_frame
- can do beautiful cross-shaped pupils, but only in close-up/portrait/eye focus pictures
- was an attempt at finger frame. didn't go well Might still be useful for extremely pretty portraits.
nahida_weighted
nahida_weighted was weighted for higher lewds while still retaining outfit fidelity. just apply standard nsfw tags as needed. (and an attempt at higher cross-shaped pupils in full body, but it didn't work)
Here is the "benchmark"
# model
AbyssOrangeMix2_hard.safetensors
OR
anything-v4.5-pruned.ckpt
# Seed
2367322303
# Size
512*728
# Sampler
Euler a; Step=30
# prompt
masterpiece, best quality, {mandatory}, 1girl, solo <lora:xxx:0~1>
# negative prompt
lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts,signature, watermark, username, blurry, artist name, (worst quality, low quality, extra digits)
Link for the models: AbyssOrangeMix2_hard; anything-v4.5-pruned.ckpt