You can use two of my models for online generation on Tensor Art:
Illustrious version 3
NAI version
But if you want to use this model here on Civit, then consider bidding some buzz to make it available on the generator!
I will bid all the buzz I get from all of you using the model! (If I remember.)
Current bid: 750 buzz on NAI V-Pred
NoobAI version:
I wanted to be part of the cool gang and hop on the NAI v-pred craze. Here's what I cooked up.
As with any of these models, use Euler or Euler A samplers. Comfy user should use ModelSamplingDiscrete and set it to v-prediction. I'll add more info here as I continue to research but for now, resources.
You can find CSV files filled with tags the model understands here. Both danbooru and e621. Which... just means prompting like you would an Illustrious model.
Illustrious LoRA models can break contrast and saturation, but they should work. Please report any issues you find.
ComfyUI users can use the example workflow here. Just drag the image into comfy to open the workflow.
General guide for using NoobAI models effectively.
Illustrious version:
Triple model merge. It's a first merge attempt, but it came out alright I'd say. Understands Danbooru and e621 tags.
It's an EPS model but it seems to work fine without EPS sampling, but I'd still recommend you use the correct settings for whichever tool you use. (ComfyUI/A1111/etc.)
I do not use A1111, but for ComfyUI you have to use the node ModelSamplingDiscrete or an equivalent custom node and set it to EPS sampling.
Use Euler A sampler, ~25-40 steps and ~4-8 guidance scale (CFG).
Positive Illustrious prompts:
(subject, {1boy/1girl/2boys/2girls/etc.}, artist/style).
Describe image with natural language. {A portrait of ...}.
(tags, quality tags).Negative Illustrious prompts:
bad quality, worst quality, worst detail, sketch, censor.Pony version:
Test model, I wanted to see if merging two of my favorite models would increase prompt adherence while still keeping its imagination.
It has a bit of a watermark problem, but they don't seem to be too intrusive and are usually easily fixed with a bit of inpainting.
Positive Pony prompts:
score_9, score_8_up, score_7_up, rating_tag, BREAKNegative Pony prompts:
score_6, score_5, score_4, etc.But please experiment, I'd love to hear your thoughts!
Description
FAQ
Comments (14)
Can the noobai be used with webui forge?
It should! I haven't tested this with Forge specifically but V-Prediction should automatically enable when running it. You might need to change sampler and scheduler settings to get the results you want though!
This is by far the best model I've used, loras work really well with it. Only bad things I could say would be the watermarks appearing pretty often and trouble with certain facial gestures. (Took a few tries to get a decent wink)
9/10
Heya! Thanks for the kind words and for taking the time to comment ❤️. I was just curious which version it was that struggled in your experience?
@Zachisalreadytaken V3, to be fair I was using 2-4 “by artist” in the prompts so a watermark was bound to appear. Aside from that it’s an S-tier model.
Yep, aaand hes done it again. Not just the best eps model version for me, but also the best v-pred model version now.
Have "only" tested a few days now, but so far its just amazing. Before this, i went from the eps version of this model to StableMond, which is a VERY similar mix to his one, also noob v-pred based, but i can already feel a few differences, even if it looks very close overall. That one was already just so nice for me, feeling the improvements of understanding, concepts and lighting, but still having most of the visual quality of good IL eps mixes. And so does this one, maybe even a bit more? Overall ill still test a few things between both these models, imo theyre just the 2 most comparable right now, and its likely just a matter of preference for small things which will make one better over the other for many people.
(What sadly isnt really comparable, from my tests, is the "normal" Indigo v-pred version, which also came out right before this one. That one seems to work quite a bit differently, with different training data? Ill look into that more too again, but so far it doesnt come close to this one.)
I'm glad to see another comment from you, and I am happy you like the model. Whenever I do have time I'll likely continue working on the NAI version like I did with IL, even though I'm quite happy with the way it is now. It doesn't have as much of an issue with watermarks/signatures as the IL model does, unless artist tags are used.
Thanks for comparing the models, by the way. I don't have much time to compare models even though I'd like to, so it's nice to save some time when models are recommended. I'll definitely try StableMond! And if you have other suggestions of models you like then please let me know ❤️.
Really really good! it can accurately make a varied amount of art styles just by mentioning the artist, althought on some of them they'll get a watermark even if you have the watermark keyword in the negative, overall very accurate though!
This one is still the very best.
The new vpred version is amazing.
But theres a weird issue with it that I never saw with any other models.
It's generating with big black bars around the sides. It happens about 30% of the time and on all resolutions and the bars are never the same.
Heya! Glad you like the model ❤️.
That's weird, I haven't encountered it personally. Could you share your base prompt, like quality tags and artist tags? It might help me figure out why it happens and maybe I can remedy it ❤️. I have seen it happen sometimes when using certain artist tags, but not as large as 30%.
@Zachisalreadytaken Well, its a bit random, with multiple different prompts/loras, so hard to pinpoint whats causing it. I did add "border, black border, letterboxed" to the negatives, and it does mitigate it a bit, but still not perfect. :/
@foxmlg Huh, as someone using this model version for many weeks now, i dont think ive ever encountered that, or at least noticed it at all. This model very rarely puts any borders for me, maybe its because of some artist tags, or some quality tags?
This is a great hidden gem model! Highly recommend trying out this model. The best part? It's super easy to use for Civitai generations, unlike other Noobai models. Give a try and you won't be disappointed!
I’ve already said positive things about his IL. However, the Noobai version is like the BIS model for anything related to lighting. The lighting here is no joke. When it comes to black backgrounds with high contrast, Noobai outperforms most other models. While others often produce a dark (but not truly black) background, this model consistently delivers a solid, pure black backdrop.
It’s also incredibly flexible! Many things can be changed or tweaked with just a few tags. It’s very easy to use and often produces amazing results effortlessly. Honestly, it feels like this model is my drug! I keep coming back because nothing else hits quite the same.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.