Simile = comparison, likeness, parallel, similar, resembling
Note: Preview images are generated without the use of LORAs
NoobAI ver. [v3.0, v4.0, v5.0, v6.0]
⚠️ATTENTION⚠️
Check 'about this version' tab for more details..
Also, here is the list for the artist tags. There are two files in the zip file, one if the full artist list and the other is the list of artist that I've personally tested to be working and most of them are also from gay artists..
About
To be honest I didn't expect to update this quickly but I can't resist NoobAI's updated datasets. Artist tags also performs much better with less dependent on emphasis to achieve a desired effect. Overall it is an improvement of the already great model, Illustrous XL.
Tips for artist styles / characters
Similar to ILXL, just add in the artist or character tags and that's it. If you're not sure what to write for the tags, just search it on danbooru or here. See sample images for examples on how to prompt the tags..
My Settings
Sampling Method: - Euler a (less detailed and smoother textures)
- DPM++ 2M (more detailed lines and textures)
Sampling Steps: 20+
Upscaler: 4x_NMKD-Siax_200k / Remacri
CFG Scale: 3 - 5
Clip Skip: 2
VAE : Already built-inPositive Prompt:
very awa, masterpiece, best quality, absurdres, perfect quality, detailed background, intricate detailsNegative Prompt:
lowres, worst quality, bad quality, bad anatomy, signature, username, logo, watermark, artist name, bad handsIllustrious ver. [v2.0]
About
Personally I still prefer the default style of the Pony version but the prompt adherence and style flexibility of the Illustrious XL is just too good to pass up. Prompting and tagging is also a lot easier than the pony version from what I've tested. I tried to replicate the look of the Pony version but there's just too much artist styles for me to do anything without messing with the overall quality, oh well..
Tips for artist styles / characters
Illustrious is a fine-tuned model trained on Danbooru datasets, so there are a lot of data on various artist's styles and characters. So you can prompt a specific style or characters without the need of loras as long as they have substantial amount of dataset on Danbooru and are not too recent. See sample images for examples on how to prompt styles..
My Settings
Sampling Method: Euler a, DPM++ 2M / DPM++ 3M SDE
Sampling Steps: 25+
Upscaler: 4x_NMKD-Siax_200k / Remacri
CFG Scale: 3 - 6
Clip Skip: 2
VAE : Already built-inPositive Prompt:
masterpiece, best quality, absurdres, newest, perfect quality, best quality, absolutely eye-catching, detailed background, intricate detailsNegative Prompt:
lowres, worst quality, bad quality, bad anatomy, sketch, jpeg artifacts, signature, watermark, artist name, oldPony ver. [v1.0]
About
Just like with my HomoSimile model, I'm trying to get that specific look that's between 'anime' and 'western art'. I went a bit easy with the LORAs to ensure better compatibility with other styles. So if you want to use a specific artist's styles it wont diminish the results.
My Settings
Sampling Method: DPM++ 2M / DPM++ 3M SDE
Sampling Steps: 30+
Upscaler: 4x_NMKD-Siax_200k / Remacri
CFG Scale: 4 - 6
Clip Skip: 2
VAE : Already built-inPositive Prompt:
score_9, score_8_up, score_7_up, highly detailed, intricate details, detailed backgroundNegative Prompt:
score_6, score_5, score_4, score_3, score_2, score_1, low quality, deformed, bad hands, watermark⚠️DISCLAIMER⚠️
Please do NOT generate illegal content with this model. You are solely responsible for any unethical use of this model(s)
Model License
Description
v4.0
What's new:
Slight tweak in details especially in background.
Very slightly reduced results defaulting to very 'bara' characters to make it easier to generate non-bara bodies.
Minor adjustments
FAQ
Comments (154)
Hohoho ! Yes!!!!!!!!!!! \o/
😃
Has anyone tried the NAI v4.0 yet? And how does it compare with v3.0?
After using it for a while, I can say that 3.0 has a high character limit and a good physique, while 4.0 has finer details and clearer lines than 3, and is of higher quality. As she herself explained, "3.0 is more like Bara."
Therefore, if you use 4.0, it seems to be a good match for the character Lora.
@yuni623 Thanks for the insight!
little bit hard to use
babe wake up, new version of homosimile just dropped
@Kairen92 do you have a list of artist names you like to use regularly in prompts for style? Male-focused ones I mean. I can't be asked to sift through 250 pages of 99% Female-focused artists to find the handful of gay or bara artists ;_;
thanks again for your work
Thank you so much for that, it's really gold among all the waifu models
You're welcome, I love your stuff too! 🥰
Should I use VAE for this model? If so, which one would you recommend?
Nope, VAE is already built-in, just set to auto. Sorry, I guess I forgot to put it in the description.
is noobai better than illustrious
from what I understand it's a finetuned sub-model of IL and seems to do very well with realism/semi-realism.
I use the NAI checkpoints with my IL 0.1 trained LoRAs with zero issues or loss of details.
@linkster123 Yes, in a way NAI is like an updated version of IL. IL used dataset from danbooru2023, while NAI used dataset from the 2024 version, which was updated until October last year I think. So it has more data on characters and styles from newer stuff. And the Eps pred version of NAI is highly compatible with almost all IL loras so you won't miss out on all the loras if you switch to NAI..
yes.
This model is also excellent at doing women and futas. Far more original and unique looks, styles, composition, etc. It's just all around an amazing model. Thanks for your work.
Thank you so much!! I'm glad you like it..
Sorry, I should have left a comment before using your model.
Thank you so much for sharing such an amazing one :)
Lol, its fine, You're not obligated to comment. I'm just happy that you like it. I love the gens that you made btw, better than most of mine .. 😂
Great work! I find some of my character LoRAs look cuter on WAI, but more faithful with this model, or "manly" so to speak. It’s also solid with landscapes, animals, and more, and covers all my other needs, so I’m sticking with it to save disk space. LOL.
Lol yup, this model is fine-tuned for more 'bara' and 'manly' characters. So if you want the traditional anime look, WAI is better for it. Or you can also use tags like 'flat colors' or 'anime coloring'.. 😊
太好了，多亏了您的系列，才能在现在AI图像的全女游戏中树立起一支男性大旗。没有你的模型，似乎都找不到好用的男性模型了，真的非常感谢。如果出一些模型训练的教程，只要价格不太超过预算，我会很乐意购买。
Thank you so much! However, I am not an expert so I'm not qualified to teach anyone. I'm sorry..
I still prefer to use 3.0 ver to make pixelart, but 4.0 do focus more on details. BTW, I just found it can draw a lot of pokemon without lora.
Yes, NAI has data from a lot of characters which also include Pokemon. So you can just prompt most pokemon without the use of lora, especially the more popular ones..
As someone who generates both bara and non-bara your checkpoint has been such a blessing. To be honest i just wanted to write my thanks, because im not seeing any noticeable differences in 4.0 😅 but i hope this and 3.0 can co-exist on the site instead of being closed due to bidding system 🙏
Thank you so much!! v4.0 has a few small changes from v3.0, most notably increased detail in the background and also veeery slightly makes the characters less 'bara' by default. Which isn't an issue for me since I like beefy characters, but for people like you who like other body types, they might not want that 'bara'-ness lol. And theres also a few other minor stuff but overall, I try to maintain the fine-tuning of the original because I don't want to risk ruining what makes v3.0 good. Ultimately, it's up to you if you want to stick with v3.0 or v4.0, whatever floats your boat lol..
Firstly, I just generate for funzzies and to do something I find very interesting. Regardless, your checkpoint was the first that completely captivated me and I have stuck with it up till the present day. It's just too good, and just too perfect to not keep trying and I will always recommend it. Keep on doing what, you're doing and I can't wait to see what's next:) Thank You and You're Amazing!
Thank you so much, you're too kind .. 😭
Really fun model to play with.
Thanks, I'm glad you had fun with it .. ☺️
@Kairen92 Sorry if I've been spamming too many posts, I'm just starting to learn how to use this tool so every time I get a result that I think is neat I want to show it off.
@Setnakht Not at all, please keep posting.. It will provide other people with examples on how to generate them..
Personally, I still prefer the 3.0 version. 4.0 may be a tad bit more stable in some aspects, like backgrounds, but in general, the most noticeable difference as mentioned by others is the "bara" effect the two versions have. 3.0 has a more prominent and intense "bara" effect, making the characters look a bit more "masculine" and bigger in muscle, compared to 4.0, which just has a bit less of that effect. At least that's what I noticed when generating with the same prompts for both versions, and I still prefer 3.0 because of the results, because I do prefer a bit more exaggerated "bara" style. Feel free to correct me though, this is just my own observation, and I hope 3.0 remains as a choice in the future.
Honestly yeah, when I made the v4.0 I've already gotten a few messages from other people that said they have a harder time generating slimmer characters with 3.0, that's why I tweak the 'bara'-ness for 4.0 a bit. But I kinda regret it a little now since I also like baras. Maybe someday I'll make v5.0 more bara friendly again..
Also, just a tip. You can still get exaggerated muscles with 4.0, just gotta add emphasis to the prompts like (huge muscles:1.5), (hyper muscles:1.5), (hyper pecs:1.5), etc..
@Kairen92 Haha, actually it's great to have control over level of muscle or bara in my generations. As it's easier to put on muscle in SD than removing them
@saka83838383 Aww I'm glad you like it. It's hard when you listen to people's feedback and you try to improve it but then other people complain about it. It's like you do this, some people don't like it, you do that, others don't like it. There's just no way to please everyone. I think if I ever make another update I'm just gonna do what I want lol..
@Kairen92 haha, you're too modest, I would talk about it all day if I ever made a model like this. And it's just very common to have different opinions over any works anywhere. The best one is always to do what you like and you won't regret.
@Kairen92 I get that, I just hope you didn't see my comment as a complaint (not necessarily saying you're talking about me, just clarifying in general), because I'm not complaining, I was just noting down the differences and that I don't see them as different versions but just different in style, and that they should remain as a choice depending on someone's own preferences! If you're referring to others, yeah, it's kinda sad, and you should just do whatever you want, if you think about doing another version another time.
@Venomee nah I know, you were nice enough so I don't interpret it as a complain. But I have gotten some not so nice ones especially in the DMs. Like full-on Karen mode demanding why my models don't work for them, why I made these changes, why is it not available for generation, why is it not on tensor, etc. Like I'm doing this for free man, leave me alone .. 🤣😅
@Kairen92 that's crazy 😭🙏 I just hope both versions remain as a choice so there aren't as many complaints and both parties can be satisfied
Works really well with https://civitai.com/models/264290?modelVersionId=396134 especially the watercolor and impressionism modes
Recently just moved to ComfyUI, but still stick with this model. No need for heavy use of lora, but the user can still easily have a grasp of the style of the image seems like a key to this model. And I'm just wondering how you can select those suitable artist style beyond the link you shared in the description. It's a hell lot of artists there, and a huge database. Any tricks? Thanks.
First of all, thank you and I'm glad you like this model.. ☺️
And yeah, it is a lot of artists, even with my personal list. Honestly my best bet is to check out my stuff and see if there's any specific style that you like and check what artist tag I used. I usually mix about 3-4 artists so that the artstyle feels 'original'. But I try to group together artist with similar style or 'feel' like for example paninterly, flat 2d, manhwa style, western bara style, etc.. So if you find an artstyle that you like you can check the artist tags in the prompt one by one and see which one you like..
I wish there was an easier way to organize these artist tags but there's just too many .. 😅
@Kairen92 haha Thank you so much for sharing your exeperience! I've been checking on your prompt for sure, that's my way of learning new artist style. Looking forward to a better way to organize artist tags in the future, if someone could come up with that. Lol
Since a lot of people are complaining about 4.0, I'll add my opinion.
When I saw 4.0 dropped, I quickly grabbed it, and my results were much better than V3!
My guys look much better.
Example of a LoRa I am working on, same seed, prompt, etc:
v3.0
https://i.imgur.com/mSyimOK.png
v4.0
https://i.imgur.com/eXb4EaE.png
He is not supposed to look as young as he looks on V3, his look is more accurate on the v4.0
Maybe I am not skilled enough, but I always struggle with my characters looking too young on my pictures, I need to rely on a lot of adult positive tags and young on negative, at least now it's easier, so I am happy
4.0 is better, the backgrounds are more detailed and I appreciate the less biased default style.
I don't notice a marked difference in prompt adherence or knowledge. Maybe more of a 3.1 in that regard.
Thank you!
Thank you for sharing! Could i use it in ComfyUI?
Of course, just make sure you follow the recommended settings.. :)
i love this
I recently been studying in Comfyui. Does anyone have a workflow they could share?
I don't understand Noobai, I get images that are just a grainy mess. Are there custom nodes i should get to make noobai models work? I would appreciate the help.
I don't know if this is the case or not, but someone DM-ed me saying that they get weird output when they used clip skip = -1 or 1. You need to use clip skip -2 or 2..
Why it’s the checkpoint down D: omg, my horny brain it’s going to explode without it ;-;, somebody fix it plsss
It seems that Civitai is having issues with their site. We'll just have to wait until they fix the problem. I know you already DM-ed me about it but I'm commenting just in case other people are wondering..
Real.
this is an awesome checkpoint im so sad that it is down as well
Yeah me too .. 😔
I already contacted Civitai, still waiting for their reply. Hopefully it won't be too long..
Regarding the 'checkpoint being taken down from the generator'. I have received a reply from Civitai and this is what they said:
"Hi Kairen92,Thank you for reaching out to Support! I apologize for the inconvenience and any concern this has caused you. Unfortunately, your checkpoints weren't in the top 200 bids for our Auctions system, which is why they're not currently enabled for on-site generation.
Each auction cycle lasts a week and bids can be placed to secure the 200 slots available for checkpoint generation. When the bidding period ends, the checkpoints that placed in the top 200 will be enabled for on-site generation for a full week, where that selection may change based on the next auction cycle. (Note: The auction rollover period always occurs at 12:00 AM UTC on Fridays.)"
So basically the checkpoint didn't make the top 200 because there was no bid (apparently). So for now it's temporarily down until the next auction cycle. I already placed enough bids to ensure it makes it the next round.. 😓
This is homophobic, and on pride month!!!! What the hell Civit?!?! 🤬
@Detoxic Exactly!! Of all the 200 models on that list its the gay one that got removed? Yeah sure Civit..
I was pretty sure I had a bid on it, but maybe it was the previous week and I'm misremembering...
@barairoAI If you place a bid on it AFTER it was down, then it doesn't for this week but for the next one..
@Kairen92 I thought I had put on a 1k bid early this week, but maybe it was the week before? I couldn't find the notification for it.
@barairoAI Idk, I find it weird too that not only v4.0, but v3.0 also didn't make the top 200 even though there are hundreds of models that are considerably less popular than these. It almost feels like it was removed on purpose.. 🤷🏼♂️
I quite clearly remember both models being pretty up there in the polls leader board, but oh well
@Slippy21405 Yeah, interesting isn't it? Almost like it's being targeted specifically. Honestly maybe I'll just change the preview cover of this model to a loli character with her boobs sticking out. Maybe then it will get equal treatment as the other straight models..
@Kairen92Have you had a bid on it? What the hell it didn't make the top 200? It was so good!
@kaminggram You mean before? I think so, if I didn't I'm sure some other people did. But apparently it wasn't enough.. 🤷🏼♂️
@Kairen92 You're responsible for that. We tipped you but I didn't even know the bidding system until now.
@kaminggram Whoa there, I didn't even know about the bidding system either, the model was up even before these auction bullshit were even a thing. Also tips are optional, not payment. I don't get paid for anything, I'm doing this out of my own time and money. So don't go blaming me for this..
Also OTHER people besides the creator themselves can bid for the models too, so you can do it yourself even if I'm not on this site..
@Kairen92 Your model is simply outstanding! I'm really impressed by it, and if it weren't for my AMD GPU, I'd be spending half the day generating images with it. It literally creates the kind of artwork I imagined when I first joined this site. Unfortunately, the site hasn’t been doing too well lately, to say the least, so I’ve set a goal for this summer: to buy two 1TB drives and download all my favorite models while the site is still up and running.
@Zignature Is there any other sites? This civitai site is not progressing at all!
@kaminggram Tensor? But let me tell you, I don't like UI of this website personally at all!
@Zignature First of all, thank you. And secondly, good idea. At the rate this site is going, I wouldn't be too surprised if it gets shut down. They're already beginning taking down NSFW models to make the site more investor-friendly, so it's only a matter of time before mine got removed too. A lot of creators have jumped ship and went to other sites and tbh, I'm inclined to do the same ..
@Kairen92 @Zignature What do you think about this site now? How's the future gonna be like?
@kaminggram Very bad. I mean they have always leaned more towards 'straight content' since forever but now its even worse. And most gay models are very nsfw so most of them got suppressed under the algorithm. They are basically trying to get investors and credit card companies and stuff on the site but since its very nsfw, they don't wanna invest. So in the future this site might not even allow nsfw stuff anymore..
@Kairen92 I think so too NSFW models are gonna get censored and only safe images will be allowed. But other sites seem to be on the same trend
@kaminggram Yeah but sites like Tensor, you don't need to do this stupid bidding war to get a model on the generator, you can just use it immediately. Plus, you can still make NSFW stuff, you just can't post them on the site..
@Kairen92 This bidding system is apparently not a good idea. You can definitely see what the investors and managers are thinking. Can Tensor also generate best quality, masterpiece, absudres and semi realistic images?
@Kairen92 Is Tensor good at generating best quality and same quality like here?
@kaminggram The quality is more or less the same, but there’s one big BUT — the site has a feature that allows models to be directly transferred through CivitAI. And I’ll say right away, I really don’t like this function, because someone already stole half of my models that I created on CivitAI using it. And if this happens to you, you’ll have to contact support and ask them to transfer the rights back to you — to your own model, XD.
I hate how they handle these checkpoint system. Pay to win is the name of the game.
@kaminggram i got better results using tensor, the problem is that their free version is very limited
UNBELIEVABLE! I finally find a model that prioritizes HOT Homoerotic men instead of random generic anime girls and Civit shots it down!
I unironically call it Homophobic Move
Yeah I just submiited bid to help you out.
And then they have the nerve to do that dumbass little corporate rainbow logo thing while they tell us that one of the most popular checkpoints on this site didn't make the top 200? Please!
@Zignature Yeah... People on there need to at LEAST ask...
I have bid this model, its 2nd now, hope going greatest again.
I didn't understand or ever use this function before, but I made my first bid just now. Guess I know what I'll be saving my usable buzz for... Thanks to OP for the work in putting this model out there, hope it can be restored soon
To all users of this wonderful model, please download it to your computer or at least to Google Drive. This site has a very toxic tendency to remove certain models without warning and even ban people without any clear reason. By doing so, you support the creator and ensure you won't lose access to your favorite model.
For every dear user, it's dangerous that this site would prevent users from using certain models or checkpoints. It's suggested to download our favourite model to our computer, enabling us to create what we dream of.
If you don't mind me asking how do you run generations through your computer and not civitai
Make sure you're posting your own generations.
It's sad that this checkpoint isn't available for, likely, a whole week now to all users. I use this checkpoint the most out of anything else, and the bidding system sucks for this, because obviously, when people get used to the checkpoint being on the site, they forget they still have to bid so that it is kept for the next week. Sadly, I don't have any available devices to be able to generate locally, so I really just depend on the site, and this makes me not want to generate anything for a week, before seeing if this returns.
I was also preparing to post now that I had actual free time, which sucks. For anyone that bids, please consider both 4.0 and 3.0, as they both offer really good and different results stylistically. I prefer 3.0, but still, both deserve to be available to users who want to use them. I hope this returns soon, it's a good checkpoint. 🙏
Man, I hate the bidding system on this website. It's so counterproductive to its growth and to people's enjoyment. I love ending my week using this checkpoint for generation and coming to see it down and out is just disheartening and frustrating to me, and probably to all who use it to create. They really need to change this bidding method because it sucks. I always try to add my bid but even with that, if you get outbid, it's just gone. Ugh, I'm sorry for all of us who are effected by this horrible situation.:(
TLDR, I think I'm gonna study up on how to use ComfyUI with this checkpoint just so this site doesn't have a chokehold on me.
FREE MY HOMO 💔💔💔 DURING PRIDE MONTH TOO???
This is such an important model for merging out female bias in so many illustrious/noobai finetunes.
I really hate how the entire checkpoint is down. I’ll try to contribute to the bids a bit, in hopes that we can get this back to normal 🙏
While I do really appreciate everyone's contribution, please keep in mind that the model only need to be in the top 200, the top spots are only for the model to be featured in the homepage of Civitai and that is pointless to me. I won't even get the buzz used for the auction. So please save you buzz for your gens instead if the model has already met its bid requirements. But seriously you guys, thank you for all the support!! 🙏🏼🙏🏼🙏🏼
TL;DR for anyone confused - apparently it's the website being stingy, not the creator. Don't DM him about it, he's aware and is incapable of fixing the problem.
Anyway, does anyone have an alternative checkpoint similar to HomoSimile XL?
Thanks in advance. 😭
Been trying some other checkpoints. There are some decent alternatives but this one is just very unique, hopefully it comes back. Check out my profile if you guys want to see which ones I use.
GUYS THE MODEL'S BACK!!!! :D
Yay
Plays rpdr s8 video of Naysha Lopez returning to the competition: bitcheeeeees i aaaaaam basaaaaaaaack!!!! 😜🤩🌈
💀💀💀
@baraisgreat lmao
We could finally spend some time with our boyfriend and husbando now,let's enjoy their penises and anuses again
How long is this open for use?
Hopefully for a very long² time .. ☺️
@Kairen92 🙏🙏🙏
Phrasing, Sailor 😏
I’m so happy it’s back 🥹😭. It has been so long since i last touched this. Thank you so much 🙏 now i can finally sleep in peace
Cock's back on the menu, lads!
Oh yes, the Franks and Wursts are back for cooking!
Lmao.. 🌭🌭🌭
This Model is Gay Men's BEST FRIEND! In a sea of Models that have a disturbing priority on anime girls
10/10 Reccomended
Nice
loooove this model
Excellent model! Love the results. Great work!
Thank you .. ☺️
Excellent model, it's great to have someone putting effort in gay models, which are a minority. There is something I'm missing: bottom/taker pov. I know there are some loras for this, I tried, but the quality is not great. I think that a good (inclusive) gay model has to include the different povs.
Not sure what you're trying to do specifically, but try tags like "2boys, anal, male penetrated, pov crotch, taker pov, etc.."
@Kairen92 Thanks for your response. I want to generate images from the bottom pov, I've been trying this lora, but changes the style too much: https://civitai.com/models/926739/concept-takerbottoms-pov-illustriousxl
and if I reduce the strength, then stops working.
I tried your tags, without loras, and I get top pov, as usual.
@acgc99 I tried the tags and it seems to work just fine for me:
https://civitai.com/images/85047190
I think you might have used a tag that is top-focused so the AI tries to correct itself and gave you a top pov. You can DM me your prompts and I'll check and see which one is the culprit.
Also keep in mind that the internet is mostly 90% hetero and 10% homo, so most images in that particular pov is just females riding on a dick and its from the male's pov. So even if you used the tags I've given and with loras, it might take a few tried to get it right. Urghh straights.. 🤷🏼♂️
This is SO good, thank you!
Quick question: How to get consistent results when generating bottom/taker's pov?
For, like, i want to create a scene where Character A doing anal with a faceless bottom, with only bottom's back & ass in the view.
But in most or even every case, although i try with relate tags, it alway generates results with Character A (especially when character A is a named character in medias) as bottom instead, which, is quite frustrating.
Is there any workaround for this case? Thanks!
Please check my replies from a previous comment from the user acgc99, just a couple of comment from yours. I tried helping them with the same exact issue as yours so check it out and see if that solves your problem. Thanks..
I would looove to use this model, can you please make an SDXL version !
Why not just use the NAI version?
Why I have emphasized short hair again, but this model always likes to grow long hair, and it can't stop the production of long hair no matter how it is described.
What are your prompts? You probably used a tag that's either female or long hair-focused..
hello everyone. I can't upload this model on RUN DIFFUSION + Automatic1111 WebUI; I upload it in the Models folder, but it doesn't appear in the application choices. Can you tell me if there is a way or if it doesn't work on this platform? Thanks in advance for the feedback
hello，Thank for your work! this ameizing!
I often encounter a problem，How can I avoid having very defined abs?
Hi, you can try using the 'belly' tag, but it likes to make the character fat though..
I just put abs in the negative prompts and it works for me.
@ILikeGarlics Thank you for your reply. Would this lead to having absolutely no abs? Or would it result in a relatively muscular but not having 7 to 9 pairs of abs
@Kairen92 Thank you for your reply. Do you often encounter situations where the characters have a lot of abs when generating content？I thank it's also possible that it's because I used Lora.
@unclepen What resolution did you generate at? If you set it too high the AI will create weird stuff like too many abs..
@Kairen92 Basically, they have a resolution of 2k
@unclepen You need to use the standard resolutions for XL models. For example:
1024 x 1024
1152 x 896
896 x 1152
1216 x 832
832 x 1216
1344 x 768
768 x 1344
1536 x 640
640 x 1536
After that, you can use hires fix to upscale it to 2k ..
Love it so much. Thank you.
Thank you so much for making these models public. I'm always amazed by their incredible expressiveness.
Do you have any plans to update this model based on Illustrious v2.0 stable?
Thanks again for all your amazing work!
Hi there, you're welcome .. :)
Right now there's no plan on making an update for IL v2 since, from what I understand, they still uses the same dataset from IL v1. Basically it means they haven't updated on the newer artist style and character tags. So if anything NoobAI still has a more expansive amount of data, which is the most important aspect for me when it comes to these types of models. Maybe someday if they'll ever make an update that includes all the newer datasets for IL v3 or maybe even NoobAI v2..
Thank you for your reply!
The reason I asked about Illustrious v2.0 stable is because I heard it had improved capabilities in interpreting natural language prompts, which made me wonder if it could offer a broader range of expression.
But of course, it's not like it has true LLM-level understanding, and as you mentioned, the advantage of NoobAI’s larger and richer dataset likely outweighs that.
Since I don't have the system specs needed to merge or train SDXL models myself, I’m incredibly grateful that you've made your models public and contributed so much to the community. Thank you again — I really appreciate your work.
I'm not very fluent in English, and I'm using ChatGPT to help me write this.
I just hope my thoughts and gratitude come through clearly.
@graphitronica Actually, NoobAI also have some natural language capabilities. Of course, using tags is still the most preferred way, but it can handle a certain level of natural language. I've never actually tested Illustrious v2.0 so I don't know how good it's natural language prompting is compared to NoobAI though.
Also, thank you for your kind words. And don't worry, I'm not a native English speaker myself so I don't care if you speak in broken English or something lol..
You made awesome model!
And I would like to ask something. Is it possible to somehow identify artist style the checkpoint used if i didn't tell it the specific one?
Hi, thanks for commenting.
Sorry but I'm not sure I understand. Do you mean you want to use a specific artist style but without adding the name into the prompt?
@Kairen92 No. I noticed that when generating pictures with certain character LoRa, your checkpiont often gives result in different styles. And only yesterday found out about artist tagging. And my point was - if it possible after generation (in case i didn't put artist tag in prompt) to identify artist style for future generation
@Lewdy_69 Ohhh I think I get it. Unfortunately no, if you don't specify an artist, the AI just decides for itself with what it thinks is the 'most suitable' one. And its usually a non-specific artstyle that can be a generic artstyle or a mix of different artstyles, so there's no way to tell what artist it uses..
@Kairen92 Oh... Well, thats a shame :(
OK, and if I go the hard way? You added two txt files with artists to the checkpoint page. One with about 800 artists (I assume the ones that your model was trained with?) and 34k (oh god; all the artists from danbooru I suppose). Maybe it's worth looking through the lesser one (if i'm correct ant those are the ones that your model was trained with) to find similoar style? 'Cause the result of generation was really good one.
@Lewdy_69 Yeah, I wish there was an easier way too.
Yes, there are two files. One is the full artist list from NoobAI, but most of them don't really work. Some do but the changes were too small to be noticeable. And the other one is my personal list of artist that I've tested to be working. Most of them are also from gay artists or artists that focuses mainly on male characters, but there are some non-gay artist in there as well.
You can either test them one by one and see which style you like or you can check the stuff I posted for more examples since I like to mix the artstyles from time to time. I try not to use lora because they tend to affect the artstyle so what you see is what you get. Good luck!
@Kairen92 Thanks!
Likewise, I would love it! I should dream of balding mature men as sign of virility in art.
Details
Files
Available On (4 platforms)
Same model published on other platforms. May have additional downloads or version variants.