I've been messing around trying to figure out LoRA training recently and decided to train a style LoRA on the original Marsey Telegram stickers.
Should more flexible than MarseyGen since this is trained on a newer uncensored model, but it's booru tag-based so it's not as easy to prompt.
Recommended Settings:
Requires a UI that supports SDXL V-Pred models such as reForge/Comfy UI
Model: NoobAI-XL V-Pred-1.0 (may work ok with other NoobAI based models)
LoRA strength: 1.0
Sampler: Euler A
Scheduler: Beta
Steps:
30CFG Scale: 5
CFG Rescale: 0.5-0.7
Zero SNR: On
Use the following tags to generate images of Marsey:
domestic cat, white fur, orange fur, black eyes, striped tail
Full example prompt:
masterpiece, best quality, marseystyle,
solo, domestic cat, white fur, orange fur, black eyes, striped tail, sitting, :D, blush, black background
<lora:marseystyle:1>
Negative prompt: worst quality, low quality
Some example images (with metadata from reForge UI):
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
Hon hon
Jump in the discussion.
No email address required.
More options
Context
More options
Context
was about to complain but realized i forgor to switch the model lmfao
Jump in the discussion.
No email address required.
No Marsey do not advance on Kherson they are heavily fortified you're gonna die! Marsey! Marsey!!!
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Marseygen uses SDXL as well, so both would share the same underlying model. These are surprisingly nice for a LoRA on the original sticker pack, though. Marseygen is a full finetune over ~200 images which I thought would've put it much further ahead of this. Being any challenge at all to prompt makes it infeasible for a bot unfortunately, only like ten people would bother learning
!sd Marsey wearing a beret and eating a baguette
Jump in the discussion.
No email address required.
can i get a tutorial, i keep getting iffy outputs like this
Jump in the discussion.
No email address required.
More options
Context
Marsey wearing a beret and eating a baguette
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
!sd Marsey posing sensually, licking her lips
Jump in the discussion.
No email address required.
Marsey posing sensually, licking her lips
Jump in the discussion.
No email address required.
AAAAAGH
Jump in the discussion.
No email address required.
!sd Marsey walking on her hind legs
Jump in the discussion.
No email address required.
Marsey walking on her hind legs
Jump in the discussion.
No email address required.
!sd marsey getting brutally shot
Jump in the discussion.
No email address required.
Marsey getting brutally shot
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
!sd Marsey walking on their hind legs
Jump in the discussion.
No email address required.
Marsey walking on their hind legs
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
Kino
Jump in the discussion.
No email address required.
More options
Context
More options
Context
!sd Marsey wearing a North Face tracksuit.
Jump in the discussion.
No email address required.
Marsey wearing a North Face tracksuit.
Jump in the discussion.
No email address required.
!britbongs !hibernians !chuds POV: you're about to either get stabbed or begged for spare smokes.
Jump in the discussion.
No email address required.
I know that fricking feel. Everyone thinks I must be a stoner or something.
Jump in the discussion.
No email address required.
More options
Context
!sd chav marsey
Jump in the discussion.
No email address required.
chav Marsey
Jump in the discussion.
No email address required.
More options
Context
Chav marsey
Jump in the discussion.
No email address required.
Oh goshdarnit
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
Yeah definitely not practical for a bot, the NoobAI model it's trained on is terrible for normal prompting.
I was also surprised it works as well as it does, it only took ~45mins to train on a laptop (after much experimentation with different settings).
Jump in the discussion.
No email address required.
Your model definitely is much more flexible. It seems like SD either wants to replicate the training data exactly or to devolve into a generic cartoon cat. Here are the images I used: https://files.catbox.moe/btu5ao.zip (captions are in the image title, and images which share the same caption have a underscore in them, like "Marsey_2.svg".) I wish the Genmoji team published papers 😡
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
Holy shit those examples are impressive, way better than the old model.
Jump in the discussion.
No email address required.
More options
Context
if y'all need a good rig to run this on, there's many unsecure comfyui instances floating around on the internet. be careful tho, some of them are FILLED with porn LoRAs.
https://www.shodan.io/search?query=title%3AComfyUI
Jump in the discussion.
No email address required.
More options
Context
Pretty cool.
Jump in the discussion.
No email address required.
More options
Context
Petite anime women Marsey when
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Kino
Jump in the discussion.
No email address required.
More options
Context
Marseyhamster is cute
Jump in the discussion.
No email address required.
I think that's a chinchilla
Jump in the discussion.
No email address required.
you're a chinchilla
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
your examples for neg prompts are shit. learn 2 prompt.
apart from that, good job!
did you use a guide for training the lora? can you link it?
Jump in the discussion.
No email address required.
I skimmed through a few guides and this was probably the best/most up-to-date one I found:
https://civitai.com/articles/3522/valstrixs-crash-course-guide-to-lora-and-lycoris-training
Looking at the metadata of loras other people trained was also useful to get an idea of what settings work for a specific model:
https://xypher7.github.io/lora-metadata-viewer/
Still a decent amount of trial and error though, adjusting the settings/captions and such and re-running the training.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I've been consuming ai slop marseys all this time?
Jump in the discussion.
No email address required.
More options
Context
It's not Christmas yet.
Jump in the discussion.
No email address required.
More options
Context
Snapshots:
Civitai Download Page:
ghostarchive.org
archive.org
archive.ph (click to archive)
NoobAI-XL V-Pred-1.0:
ghostarchive.org
archive.org
archive.ph (click to archive)
Jump in the discussion.
No email address required.
More options
Context