NoobAI guide
WIP:
Likely to change as the model is new. This guide works for both anime and furry art.
No I will not use JPGs. YOU WILL let this page load.
V-PRED
V-pred versions including EasyFluffXL are in development, some bugs exist on reForge, so I will wait with updating the guide for them.
Base Models
Recommended Merges
Noob and its merges currently come in three varieties, mainline, lightning, and vpred. The lightning models primarily come from furry merges. Vpred is still in training for the mainline model.
Mainline Noob is a bit of a mess. Due to an overtrained text encoder (at least that's what people currently believe), the 1.0 model does not strictly perform better in every aspect compared to 0.5 or 0.75. China and the anime genners blame the introduction of the furry dataset after 0.5 for introducing 30000 images of diaper porn. Funny.
I tested most of these at least a little. Things are developing fast, for the time being I will mainly gen on Zoinks, since refining a stylemix takes time and I can't be arsed to re-discover every 2 days.
Regular models:
Mainline:
Noob 1.0 (Recommended)
Noob 0.75 (Middleground)
Noob 0.5 (More creative but schizo)
Vpred:
Currently in training. Some merges have tried to merge the in-progress Noob Vpred into 1.0, but they don't seem to improve much yet.
Running Vpred in reForge:
Make sure you are on the dev_upstream branch. Normally your Vpred models will get autodetected, otherwise on dev_upstream, enable the 'Advanced Model Sampling for Forge' tab below your seed. Tick it, "v_prediction" and the ZSNR box.
Beta scheduler is currently broken. EulerA CFG++ seems to work well with these models too.
Merges:
Jinkies (Basically a full-step count Zoinks with some scuffed natural language capability) (Recommended)
EasyFluffXL (In development. Very comparable to Jinkies still)
Low-Step / Fast merges:
Lightning:
ZOINKS (Merge of ChromaXL from the EasyFluff guy) (Recommended)
ChromaXL (Classic Lodestone-ware (compliment). Currently comes in Mango and Spud flavors, don't ask me which is better. Supposedly Spud almost the same as Zoinks.)
VAE:
fixFP16
Minor bugfix VAE you can use over the regular one.
(Further advice will assume Zoinks is used.)
Prompting NoobAi
Using the examples off of Civit is a mess. The authors either don't know how to prompt their own model, or have included submissions by people who don't know either. Take the following with a grain of salt, since a lot has been gathered from community observation.
These apply to all Noob models.
1: Quality tags:
(masterpiece, best quality, very aesthetic, newest, 2024,),
(worst quality, normal quality,),
Very aesthetic seems to have an effect. Newest and 2024 are lesser known ones that work. I personally don't use normal quality in negative.
2: Artist tags:
(artist:syuro:0.8), or (syuro:0.8),
"by artist" was not used. It is not clear which one is "correct", as the model authors seemingly can't decide either.
Different merges and the basemodels require drastically different strengths on artist tags due to the overtraining! Some want them as low as 0.2-0.3, some higher. Zoinks seems to be okay with a 0.6-0.9 range.
3: Sampler & CFG:
For mainline models use 30 steps. Noob currently vastly prefers Eulers, especially ancestral, due to problems with noise and graininess.
Euler Ancestral CFG++ & Beta @ 14 steps @ 0.8-1.0 CFG (recommended)
or
EulerA & Beta @ 30 steps @ 5 CFG
Setting the CFG lower than 1.0: Open the ui-config.json in your reForge root folder. Ctrl-F and change to this.
Other useful tags:
multiple panels
In negatives if you are getting unintended Manga style layouts.
solo, anthro,
In positives is almost required to avoid filling the scene with multiples.
watermark
Probably works but I prefer to edit them out, then again my mixes never have them. (RIP Zackary posters)
4: Other hints:
1: There is 0 natural language prompting in the base model. This extends to even smaller tag associations like "green hair tips" not always working as well as on other models. The upcoming Zoink sequel is attempting to get a bit of NL back in.
2: The base model is Illustrious, therefore loras trained on it will work. Even Pony loras work to a smaller extent. Civit has an IL filter.
3: Noob is fairly competent at representing even low-tagged concepts. Just don't be surprised if your favourite Taiwanese platformer is suspiciously missing.
4: You can grab an updated autocomplete taglist csv from here
5: You can generate bases on Noob, and then refine them in Pony if you have a setup you like. I recommend doing a 1scale pass @ 0.4 denoise, and then your upscale @ 0.3 denoise. Seems to work for me. I partially do it for style, but also to get rid of the grain/blur that Noob tends to have.
Templates:
Basegen:
Upscale
1.75 Scale / 0.4 Denoise
| EulerA CFG++ Beta | DPM++ 2M Karass |
|---|---|
 |
 |
| Smoother | More detailed, but has that 'haze' over it. Only using it for the upscale avoids 'some' of it. |
FAQ
My anatomy is fucked up bruh!
Less of an issue on Zoinks and Chroma. Remember Noob Really likes following your prompt. If you tell it (ass up, handholding), then your waifu will become a contortionist.
Shit's grainy af dawg!
Try eulerA samplers. It's a known issue with the model in general. They did close to 0 dataset pruning, we ended up with more jpgs in the training data than Pony.
My inpaints got bigger halos than the average blue archive cunny!
Euler's gradient is far more smooth than DPM 2M, that means that lower denoises will behave stronger. Set your denoises lower. Alternatively set the CFG slightly lower. Soft inpainting is also a toggle that is a no brainer for this stuff. Just press it, only adds a second or two to the gen time.
What's this CFG++ shit about?
Quoting Clybius:
"Instead of using noise + denoised_prediction * step_size, we do denoised + (noise - unconditioning) * sigma
You are to use it on a scale of 0-1, (may be different depending on the ++ sampler,) it has more structural stability than regular CFG sampling, especially at low cfg values.
We're basically using the raw prediction, and taking out the unconditioning from the noise before adding more noise back onto the latent.
basically just better :P I love it because its low cfg-level quality, with the structural stability of higher cfgs "
Where's my realistic dogpussy.
I highly doubt there will be a competent realism model anytime soon. Noob and Pony are both averse to this sorta stuff, noob even more than Pony was.
Duos? Regional Prompt? Controlnets?
Duos can be raw prompted. I don't do a lot of "OC on OC" to have bleedover issues, if I did I would inpaint.
I don't do regional prompting these days, too finnicky.
Controlnets supposedly kinda work. The Noob team said they are training Controlnet models for everything, so it's coming at least.
So what is better, Noob or Pony?
Pony intentionally made itself dumber. Noob accidentally made itself dumber. Pony worked out at the library, Noob is the tard that can rip your arms off. (The answer is Noob).