Mesmera is a LoRA (LyCORIS/LoCon) trained to enhance the capabilities of Illustrious 0.1 (and it's finetunes ofc), in generating hypnosis/mind control-related scenarios.
Tested on these Illustious-based models: MiaoMiaoHarem v1.5a, NtrMix vXII, WAI_NSFW v9, Hassaku v1.3A.
Training Details:
Utilizing a large dataset of 22,000 images sourced from hypnohub, this LoRA focuses on a variety of hypnosis-related themes.
2xT4 on Kaggle, 30gb vram in total. ~15 hrs training time, 8 epochs. "I forgot the number but it's B I G" steps.
Used code from Hollowstrawberry's Google Colab repo, modified, and adapted to Kaggle by me. Kaggle repo (code is a complete mess), find the Training SDXL Lora on Kaggle.ipynb file and import it into Kaggle.
All dim alpha values was set to 24.
Optimiser: Prodigy
Scheduler: REX
Batch size: 6 (3 per gpu)
Grad. acc. steps: 1
Additionally used IP noise = 0.1; Prodigy 1.1.2 "slice_p=11" arg; --debiased_estimation_loss
Dataset:
The training dataset consists of ~20k images scraped from hypnohub, all centered around hypnosis and related concepts. For a full list of tags used in the dataset, you can refer to the Bottom of the Tag Guide section where you can find a Gist link.
Dataset was scraped from hypnohub with this prefill filtering:
-manip,-animated,-animated_gif,-voice_acted,-animated_eyes_only,-tagme,-traditional,-sketch,-meme,-ai_art,-caption,-caption_only,-fascinum,-3d_custom_girlAll images was downscaled to 1024 (bigger side) pixels by lanczos method in XnConvert tool.
My booru scraper script also gathering tags from site. After additional filtering (see below), further tagging with JoyTag (high confidence like 0.6) in the Dataset Helpers tool.
Additionally some further filtering were applyed:
"score:>x" tag, where x value is defined by me for every main tag.
Aspect Ratio filtering: for every image were calculated Aspect Ratio value (for example value for AR 16x9 will be 1.77), and images that were too long or wide (with an AR value greater than 3) were filtered out.
After filtering and tagging, tags in each .txt file was organized by this metod:
[Concept tags] = always first in the file, sorted a-z.
[Style tags] = not sorted, simply moving them after Concept tags.
[Main tags] = [Concept tags] + [Style tags]. The number of main tags is calculated. This is our keep_tokens value.
After these steps, .txt files looking like this: [Main tags], other tags. Then all pairs of .txt + .img were sorted by keep_tokens value. Finally, I got a folder structure whose name depended on the keep_token value (i.e. 1, 2, 3, etc.).
Tagging Guide:
To utilize Mesmera effectively, you can explore a range of styles and concepts. Here are list of the main tags I was focused on:
Meta:
femsub: 700 <-- female submissive. use it to specify on whom chosen hypnotic effect will be applied
before_and_after: 676
pov_sub: 557 <-- Point of view of submissive character
sequence: 539 <-- can be used either as sequence of transformations of given character or just comic-style sequence
see-through: 468 <-- contextually aware, i think...
malesub: 426 <-- male submissive. same as femsub
maledom: 309 <-- male dominant. use it to specify who is applying effects on fem/male sub.
gameplay_mechanics: 289 <-- idk what is that, but it should help with some interfaces or level bars shit i guess
femdom: 270 <-- female dominant. same as maledom
pet_play: 222
Concepts:
kaa_eyes: 700
happy_trance: 698
bimbofication: 683
hypnotized_hypnotist: 680
transformation: 553 <-- random transformations in training data, can help can make things worse
breast_expansion: 512 <-- warying breast size, depends on artist style
hypnotic_audio: 493
haigure: 480
dronification: 460 <-- warying results, general non-specific tag
evil_smile: 444
hypnotic_screen: 435
bodysuit: 417
hypnotic_app: 415
tech_control: 402 <-- warying results, general non-specific tag
control_indicator: 400
robotization: 365
clothed_exposure: 359
visor: 355
progress_indicator: 331
hypnotic_accessory: 329
lip_expansion: 326
zombie_walk: 316
hypnotic_music: 311
pendulum: 311
mantra: 304
altered_perception: 300 <-- use with "thought bubble" tag
enemy_conversion: 291
hypnotic_eyes: 279
hypnotic_gas: 271
charm_(spell): 271
empty_eyes: 270
hypnotic_breasts: 269
electricity: 268
barcode: 266
hypnotic_penis: 262
hypnotic_light: 259
confused: 259
corruption: 242
chicken_pose: 235
spiral_eyes: 228
unhappy_trance: 212
shrunken_irises: 203
glowing_eyes: 201
hypnotic_beam: 184
standing_at_attention: 182
latex: 178
self_hypnosis: 176
housewife: 167
stage_hypnosis: 160
aura: 160
hypnotic_tentacle: 154
eye_roll: 152
hypnotic_ass: 139
harem_outfit: 132
hypnotic_feet: 129
asphyxiation: 103
Characters:
erika_(er-ikaa): 510
kaa: 315
crystal_(zko): 226
hypno-tan: 162
kassidy_(medrifogmatio): 134
mrs._erickson_(zko): 59
Styles:
sleepymaid: 624
jimryu: 512
darkhatboy: 345
oo_sebastian_oo: 331
hadant: 308
mythkaz: 267
nexus_light: 253
lairreverenteboladepelos: 243
etlabsotwe: 234
polmanning: 223
zko: 214
brellom: 200
katsiika: 199
onefeefoor: 196
eroborne: 196
zephyrgales: 177
mahoumonsterart: 177
apopop: 176
shishikasama: 174
smeef: 171
faetomi: 170
dochaunt: 167
porniky: 165
djuuicebox: 164
enetheligthingdancer: 160
ryuugu: 158
brushie_art: 152
efalabrino: 151
eshie: 151
glowhorn: 151
cursedrooks: 150
myuk: 150
psyfly: 148
4headboiii: 148
b-ginga: 144
detritus: 143
medrifogmatio: 137
artofadam: 127
batta18th: 126
ghostec: 126
gmun: 125
keeper_of_pots: 125
idpet: 123
horiizyn: 122
pstash: 122 <-- not working
tomo86: 119
supercasket: 118
ameerashourdraws: 115
tunberuku: 113 <-- 50/50, dirty dataset
httpwwwcom: 112
electrickronos: 112
erocoffee: 109
abarus: 109
yumiiart: 108
wrenzephyr2: 108
foolycooly: 108
lapislazuliart: 105
zelhypno: 105 <-- 50/50, why? idk :(
supersatanson: 103
vahn_yourdoom: 101
zorro-zero: 100
yensh: 97
davidthewolfx10: 94
zombi62: 94
maozi_dan: 89
jostony24k0: 88
rosvo: 83
alerith: 82
reliusmax: 81
m4ns0n: 81
konaloid: 80
maynara: 80
sakurarose12: 79
sweetlittleneko: 78
malberrybush: 74
harvestman_here: 70
kronobas28: 68
orphan2: 68
syas-nomis: 63
zelamir: 62
nettleseeds: 61
4five1: 61
the_iron_mountain: 56
kibazoku: 55
cavitees: 55
sl33pyg1mp: 55
shozaya: 54
gerph: 50
blueparikeet: 49
drevod: 48
shieol1: 46
hmage: 45
konno_tohiro: 45
rnslivr: 44
deepspaceart: 42
singlesalt: 40
magukappu: 39
borvar: 38
glatu: 38
cuddlesword: 36
nez-box: 36
terasu_mc: 36
strwbrrychz: 35
thiccwithaq: 35
a_singular_fish: 33
hentai_man: 32
chien_vietnam: 32
alittleshyart: 30
rind_rin00: 28
slugbox: 24
Full list of tags in the Mesmera's training dataset can be found in this Gist.
Mesmera v1 Gist links:
Tagging guide: Gist.
Full list of dataset tags: Gist.
Roadmap (kinda):
So... This LoRA is mostly working, but I'm still not satisfied with several quality aspects (for example, it was trained at 512 resolution). I will likely release Mesmera v2.5 this year, though I don't know exactly when. I feel slightly burnt out by this project, and I'm not going to push myself into it any further. I will focus on small-effort LoRAs for now. (27.03.25)
After I release Mesmera v2.5 for Illustrious 0.1, I will likely follow this training plan:
NoobAI Eps
NoobAI V-pred
Pony v6
And after Pony I will mostly handling user requests:
Training for other models on the SDXL architecture: IL 1.1, IL 2.0/3.0 (hopefully), RouWei, Hassaku, WAI, etc.
I will also try to train for Flux, Lumina, SD1.5, SD3.5, and Sana.
Additionally, I will try to experiment with realistic style fine-tunes (Illustrious, Pony).
Description
First release! Not perfect, not absurdly bad either. Yaay!
FAQ
Comments (28)
Thank you so much! Especially for such a comprehensive set! Although just a slight request, could you also see if tunberuku's style is in there?
Thanks a lot for your kind words!
Unfortunately, the tunberuku style did not get into the dataset when downloading. But thanks for the suggested style, I will definitely add it to my list for Mesmera v2!
By the way, you can always check if there was any tag in the training data, in this Gist - https://gist.github.com/klinok64/1094d1b865945a6c9203cd3cafb1c327. This does not guarantee that lora has learned this tag properly, but there are still chances, especially if the tag is represented by >80 pictures.
@Scorponov Certainly looking forward to a future version or two, it's got a lot of potential! Especially since some hypnohub only stuff like zko's visor or niche hypno artists only get psoted much there.
this is wonderful, and i love it, but i hope if you could add in the dataset some images from sakurarose12 as extra pls
Thanks for the suggestion! I'll add it to my list of planned tags for Mesmera v2.
As I scroll through the pictures, I have two important considerations:
1. The number of pictures may not be enough for the model to learn the style properly.
2. Most pictures have a heavy bias towards the 'comic' tag. This means the style may either be influenced by comic structures (e.g., generating unwanted dialogue bubbles, multiple faces/characters/views) or may require the 'comic' tag to be properly activated.
Anyway, I'll do my best!
@Scorponov well yeah, since most of her art isn't related to hypnosis it's not gonna be available in hypnohub, the place to see almost of her draws (which includes both her comics and individual art pieces) it's her tumblr with a certain tag https://sakura-rose12.tumblr.com/tagged/roseyart ... but the taggs she uses there wouldn't be of any help for adding them to this lycoris. (she also has accounts on Twitter/X, bluesky,
On the other site it would make sense to simply make her artstyle a separated Lora (or other kind of resource), but well that would be an entire new proyect which means a lot of work... at least she it's a really good artist with different outfits, just like in this post i found:
https://sakura-rose12.tumblr.com/post/769581647868952576/new-commission-for-goggy
Pretty amazing stuff, i'm definitely looking forward for the next versions!
I think the puppet/human puppet concept would be a nice addition as well.
Also, if balancing both concepts and artist styles in one lora becomes tricky, i wouldn't mind just a lora focused just on the hypno concepts (without specific artist styles) if that's easier, but of course, totally up to you.
Great stuff nonetheless :)
Yeah a lot of these are specific ideas that illustrious can't do on its own, like dronification, or brainwashing visors etc, those I think are more interesting if one or the other.
nice job can you make one for IL 1.0 and 1.1?
Isn't IL1.1 On-site degeneration only?
Anyway, here's a rough roadmap:
Mesmera v2 (currently training) for IL0.1 -> Mesmera v2.5 (polished v2) IL0.1 -> v2.5 for NoobAI v pred -> NoobAI eps -> Pony v6 (? not sure) -> requests for other models.
I hope that after I train mesmera for pony, Onoma release ILv2. cause results produced by ILv1 and ILv1.1 are kinda meh
@Scorponov IL 1.1 is available on hugging face it's just not listed on civ sorry for not clarifying. tested it on 1.1 and it just well makes the image but but its great on 0.1
Works quite well and very flexible with the concepts and styles. Can require a bit of fiddling when paired with other LORAs but nothing too troublesome to deal with. I look forward to future versions!
Thanks! The next version isn't going to come out soon; I'm feeling kinda burned out :(
Plus, I'm planning to clean Mesmera's dataset, because for v1 and v2 I used just basic filtering – there are a lot of images that are just unusable or too different from the original concepts. And better tagging, ofc – I'm thinking about using Joycaption instead of Joytag. I'm also thinking about masked finetuning after training 2.5 (maybe condensing my dataset to 200-500 high-quality images, making masks for specific concepts, and so on)
All of these need some time. Maybe, just maybe, I'll start working on 2.5 in the summer and release it in mid-autumn
It's improved quite a bit, although I notice that it has a bad tendency to produce 3 girls a lot even if prompted for solo or two.
Yeah, sorry about that :(
It's because I relaxed a filter in the script that gathers images
Try using negatives like multiple girls/doms/subs
Due to Civitai's TOS update this lora will most likely be deleted soon. SHIT.
"Content depicting sexual activity while in a mind-altered state is prohibited, including;
Being drunk, drugged, under hypnosis, or mind control"
If you know alternative sites where I can upload this lora and continue working on it, let me know!!!
just don't post x/xxx rated sample images?
@sliptrick idk :/ I think my LoRA might fall into the prohibited category now since it's meant to produce explicit content. I'd be happy to be wrong though...
@Scorponov It can also produce non-explicit content, so should be fine.
(In a sane world... well, in a sane world we wouldn't have that rule in the first place, but whatever. Should be fine.)
I feel you. I got so pissed off by that, that I am seriously re-considering creating any more hypno-themed content. It feels like the whole world is out there to get rid of your kink...
DiffusionArc just launched in response to the Civitai TOS updates. You could maybe upload it there. No idea if the site will stick around for a long time though.
I don't think I'll leave Civitai entirely. In fact, I'm thinking I'll upload a multistyle-only version of Mesmera v2 here. Let's call it "Mesmera Lobotomized"
You may try to keep it as it is, just "rebranding" it as "hey guys, that's just style collection, and here is a link of tags i used during training".
@kanareika1 Yeah, will try that
Maybe I'll just remove any imgs that are X/XXX as other comment suggested.
@Scorponov Do that, but you can also mirror it somewhere. I think Tensor.art is a popular choice, though I'm not so sure on their policies and how easy or hard it is to train and upload models. Seaart doesn't allow for adult content anyway. I want to see this continued so if you do choose another place then maybe link it in your profile?
Also Huggingface is a popular backup site. Not much in terms of engagement but handy to store files from what I've seen.
To clear things up:
I'm not panicking nor purging my presence of civit
"Leave the ship" thing is meant to warn those who use Mesmera for upcoming changes (lora removal/lora migration to other resources/Mesmera lobotomized). It's just a big name for attention, nothing else
Thank you for such a great model, I enjoyed my time playing with it, even if that time is over. If you find another site please let us know.
Thank you for making this model regardless of how not nice the folks at CivitAI are.. hope you can find a new and better platform :)