Four new ways to get seen and get paid: sticker placement, remix galleries, cosmetic packs and collab collections. Plus free placements on every image, so you can join in without spending a thing.

The Creator Collab Update

You can gift a membership to someone else, or buy one with Buzz instead of a card. Here's exactly what each one gets you.

Membership Update

Sell a set of cosmetics as one listing at one price, instead of one piece at a time. Two to twenty items, any kinds mixed, other creators' work included.

Cosmetic Packs

We're a Day-0 launch partner for MiniMax H3, a next-gen open-weights multimodal video model - and it's already in the generator. No waitlist. Open weights land in a few days.

MiniMax H3 is live: open-weights video, generate now

After creating a realism Lora for PonyXL, I decided to create one for illustrious, as it seems to be able to create an even wider variety of characters than pony, and it seems to be a lot better at prompt adherence. This Lora should be able to make realistic images in a variety of illustrious models. I recommend you check the release notes each version of the model for what it can do.

I trained the model locally on the same dataset, but this version is much weaker in terms of the style as I didnt know the optimal paramters to set it o. It doesnt seem to do much of anything on pure anime Illustrious models outside of make the style slightly more realistic and detailed, but not photorealistic. However on NoobAI derived models, realistic models like reillusm, and modles like WAI, it seems to improve realism, in particular on models that have a small realistic image dataset but arent fully realistic (like NoobAI). The recommended trigger word is "source photo", but other realism takes help to bring out the style more. I recommend a strength of around 1.0, but on models that are already tuned for realism, you may need to turn it down to 0.9 or 0.8 to avoid distortions. I also recommended putting "anime,comic,plastic skin,overexposure" and the like in the negatives. For the sampler, I recommend Euler or Euler A, as DPM++ seems to produce distorted images, though maybe its due to my clip skip being set to 2. Euler A seems to work best at a CFG of around 3.

Realism Enhancer Beta

This version was trained on the same dataset as the prior one, but with captioned images and slightly more epochs. It seems to produce significantly more consistent realistic images regardless of sampler and model, but works best on DPM2A with a CFG of 2.5 and a step count of 18-28 or with Euler A at a CFG around 3 with 30-35 steps. The trigger is simply "source_photo", but adding a lot of positives and negative seems to dramatically improve the quality. Check the prompts for each image for more info, it seems using the recomended keywords for NovaReality and Reallusm Checkpoints also helps with this LORA.

Beta 0.5

Updated version trained on an expanded dataset using civitai's training service. This version produces significantly improved backgrounds,especially natural ones, and better animals though its slightly worse at making vehicles. The realism for some characters is also not as strong, but the overall realistic effect of images is improved. I recommend using DPM2a as the sampler with about 28-35 steps and a CFG around 2.5. The trigger word is just "source_photo", but adding "realistic,photo_(medium) can also help on some NoobAI based models.

Beta 0.4

Improved illustrious version of the model, trained on a slightly different data set using civitai's trainer. This version seems to have significantly improved the background quality of images, though it can take some finagling with the settings to get a realistic results. It seems to work best on DPM 2A with clip skip of 1 at around 30 steps with a CFG of 2-2.5. The trigger word is just "source_photo", bit adding "photo_(medium), realistic) on NoobAI derived models can help enhance the result further.

Beta 0.3

I altered the training settings on the model a bit and now it seems to produce significantly better backgrounds, especially for indoor environments, Outdoors and natural environments are also significantly improved. I recommend changing to a clip skip of 1 when you use this lora as clip skip 2 is more anime focused. The most effective sampler settings are DPM2 A at a CFG of 2.5 at 23-30 steps, with a LORA strength at around 0.8. The trigger word is "source_photo", and for some Illustrious models such as NAI, "photo, realistic" may also help improve the quality.

Beta 0.2

Reallustrious (Photorealism for Illustrious XL)

physical violence

weapon violence

wide hips

revealing clothes

downblouse

convenient censoring

pg-13

corpses

suggestive

oral invitation

pg13

sexy

huge breasts

thick thighs

sexual situations

male nudity

disturbing

male swimwear or underwear

female swimwear or underwear

partial nudity

undressed

female nudity

breasts out

exposed female nipple

breast out

lingerie

male underwear

hair over breasts

female swimwear

gigantic breasts

no panties

graphic violence or gore

covered nipples

huge butt

strapless leotard

sitting on face

emaciated bodies

one breast out

female underwear

nude

nsfw

graphic male nudity

adult toys

illustrated explicit nudity

nudity

graphic female nudity

hentai

futanari

porn

sexual intent

genitals

peeing

vore

oral

sexual activity

anal

blowjob

dildo riding

incest

hanging

hate symbols

nazi party

white supremacy

diapers

scat

self injury

hate speech

urine

extremist

child on child

latex clothing

swimwear

bukkake

fellatio

cumshot

implied fellatio

eat_cum

cumdrip

cum in pussy

cum on face

after fellatio

cum on hair

cum on body

cum on tongue

cum on hands

cum in mouth

triple fellatio

autofellatio

fucked silly

cum on pussy