Well, it is "fictionalai"

@FunkMonkerson .......get out lol

kkkkkkkkk

Pony v7 is only available in discord for money, bruh

2P2 I hope you understand that it was a joke. Did you see it? Is it good?

itisKos Nothing that's behind a paywall and never discussed is good

touch grass

@jack286 sybau 💔

I have decided. I made my model. If you like ponyv6 I think you should check it out. It is very close to base but is easier to operate then basev6. If you download you can see how I balanced it and I also have loras used I pics on page that I find are good candidates to pick from. Too many to list here. Its called raver girl - absurdly balanced pony.

If you're generating locally - e.g. on Cloud like RunPod then it might be caused by VAE incompatibility. During generation the image looks normal but the it displays with white dots and blurriness etc. I then just switch from automatic VAE to sdxl_vae.safetensors which is basically one click adjustement depending if you use preinstalled template e.g. A1111 on RunPod

fucimansonofwisdom756 nevermind im just a dummy and didnt realize the actual base model was bumped off the top 200 and civitai switched the one i was using before with this one. no duh i got different results haha.

fucimansonofwisdom756 thanks for the help! its actually just the base model i was using before was bumped off the top 200 without me realizing. i was just being a noob.

It is XL, run in ̶f̶o̶o̶o̶c̶u̶s̶ ComfyUI

@2P2 yes I already moved to comfyUI. Sorry for the dumb question again

doesn't matter if you run it on fooocus or comfyui. it just needs specific pos/neg prompts to work properly. And standard pony v6 model always works best with a style lora added.

"Something went wrong. There was a problem on the image generation. Please go back and try again :("
Thanks very much

Cant make nsfw lol

Nice joke, I know you actually have been using Illustrious

@2P2 Yeah, but before using illustrious, I mainly used pony but didn't want to upload it because often I just tested images or played with some anime/realism image generators before finding out that illustrious is much better in anime style.

TRUE

Хз, йопта

Астралайт (создатель Пони) слился. Он Поню обещал аж в январе выпустить, потом перенёс на конец весны, затем в июле сообщил, что осталось меньше двух недель, и с тех пор снова тишина.

@Chertilo семерку в январе обещал?

@dishaforex726 да, семёрку в январе. Прогресс-бар был заполнен на 99%. А потом вкрысу его убрал, типа и не было ничего.

@Chertilo Как говорится: надпись «Завершено на 99%» радует только первые три часа!

@Chertilo есть неприятные предположения что модель вышла попросту неудачной. И даже если так, то ауру никто развивать из комьюнити толком не будет даже с бустом в виде в7

@yuduz367 тоже наслышан, что семёрка кишит проблемами.

Тут в Дискорд-канале Астралайта, оказывается, появилась инфа, что релиз Пони 7 состоится буквально на днях. Сильно в том сомневаюсь, учитывая, сколько раз он срывал сроки. Но если даже вдруг и релизнется, есть немалая вероятность, что модель разочарует. Тем более что сам Астралайт просил многого не ожидать, плюс циркулирующие слухи о проблемах с моделью, самый тревожный из которых - очень медленная генерация даже на картах с 24 Гб VRAM.

@Chertilo Лучше бы v7 оставили на SDXL, это бы не помешало ей быть лучше v6, достаточно посмотреть на Illustrious и NoobAI

@sukablatnahui600394 он поначалу так и планировал, но потом чёрт его дёрнул выбрать говно под названием Аурафлоу, являющееся к настоящему времени полностью мёртвым проектом.

@Chertilo Странное решение, тем более если нейронка не тянет карты начиная с rtx 4070 то она нафиг никому не нужна будет, как правило аниме модели нацелены именно на массового потребителя, а как говорят у v7 с этим большие проблемы и она даже медленнее чем FLUX

@sukablatnahui600394 тем более что объективно там нет ничего, что оправдывало бы такой бесстыдный аппетит до VRAM. Если оно и в самом деле будет настолько жручим, то REST IN PISS Поне тогда.

@Chertilo На самом деле, это очень странно, понимаете? Мы все знали, что AuraFlow — отстой, и все советовали ему использовать Flux, но он наотрез отказался. По поводу проблем с лицензированием или чего-то в этом роде, как я написал в посте на Reddit.Мне кажется, что Pony V7 станет катастрофой именно потому, что он основан на AuraFlow., но, по крайней мере, у нас есть и другие варианты помимо SDXL, например Chroma

@Light_Diffusion вроде бы именно ваш комментарий, если не путаю, я видел на Реддите под одним из обсуждений Пони 7.

Аурафлоу изначально казалась крайне сомнительной по всем параметрам. Мне кажется, Астралох потому и тянет с релизом, зная, что это будет полная катастрофа, которая уничтожит его репутацию.

@Light_Diffusion В целом Flux тоже сильно избыточен для аниме модели. Но всё лучше чем AuraFlow

@sukablatnahui600394 Флюс несравненно лучше. Я повидал кучу лор на нём и все они, за небольшим исключением, выглядели прекрасно. Чего не сказать про Хрому, где с ними полный кошмар: уровень там ХУЖЕ, чем у полторашки (1.5). Опасаюсь, что с Аурафлоу, если она вообще выйдет, будет такая же беда.

@Chertilo @Chertilo Да, в любом случае ждать v7 нет смысла, оно не будет лучше того же waiNSFWIllustrious, а раз не будет то и смысла в нём нет даже в качестве альтернативы. тем более это AuraFlow которую никто не любит и почти никто не знает.

@sukablatnahui600394 база.

Sometimes too many negative prompts can be just as bad as not enough. CFG 7 and 35 steps is always my sweet spot. If it’s not that, then I am not experienced enough to know and apologize.

It's a base trained checkpoint similar to Illustrious v0.1 and NoobAI, meaning the results won't be consistent or stable unless you're not using LoRAs and fairly simple prompts. Besides that, Pony is known for being fairly rigid and difficult to train LoRAs on.

This is also why V6 XL has the note "Start with this one" as merged checkpoints like Autism Mix, Pony Realism or 2DN typically provide cleaner results.

you could try to fix hands with the mesh graphormer, or use detailer nodes with bbox / segm stuff

我在comfyui里用这个模型的效果也不好，甚至是很坏，不知道是什么原因....

@smaft sure, but what I wonder is why this model doesn't do it right away like many other models since it's at the top of the search on civitai. I've finally come to the conclusion that it's an older model (requiring more tweaks), downloaded longer than others, and that it's not a poor configuration of mine.

Eh, it'll need to be generationally better than pony 6 to warrant the wait, especially with noobai and illustrious being as good as it is already.

@MysticFrostwind Is there even any point in using the PONY checkpoint? Or is it pointless? (._.)

@Oskarjasz Meh, there are some good pony finetunes for things like realism that illustrious doesn't do as well, but most things I'd say NAI/IL just outperform Pony v6, though again the latter was great for it's time.

@MysticFrostwind Thank you for your answer❤️

С таким же успехом можно ждать прилёта инопланетян или пришествия Христа.

the only news is the same since months. apparently there is a ponyV7 somewhere on purplesmart, but that's it. nothing more, no news from the blog, etc...

@daruthin on discod he says other words

Everybody ask "When is v7?".

Does anybody thought asking "How is V7?"

I don't think so.

never, illustrious made pony completely irrelevant anyway

use runpod (rented system)

Try fp8

install forge ui i have laptop 3050 4gb and forge ui is the only way to go takes around 40 sec to 1min depending on settings to gen image

Hey I have a suggestion. So the problem may be how your settings are configured. I’d recommend using fooocus. You can look it up on google “ fooocus GitHub “ and follow the instructions to set up the gui. There will be a pony preset available if you check the advanced checkbox. It will generate good results by default and you can fine tune it by adding loras if you want also

To get the model to work as intended you need three things, the VAE recommended in the info placed where your system needs it, setting the Clip Skip to 2 in the generation settings, and using string "score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up," before the usual prompts.

Missing any of those, especially the Clip Skip, will severely hurt the generation.

Because the model is bad! You have other, better models

@2P2 Oh! I have downloaded a lot of other models now. Their results are indeed very good, and I've also found some that I like to use!ξ( ✿＞◡❛)

@HarryShezar Oh... I actually did have the VAE downloaded and the steps set up. I think perhaps I missed a keyword or two in the prompt? Thank you for your answer, I will go and look into the reason!( • ̀ω•́ )

@jbulman00401 Oh! Thank you very much for your suggestion, I will go and give it a try. Thank you!٩(๑•̀ω•́๑)۶

I encountered the same problem, I realized that the model is 2 years old, maybe it was good back then.

Your probable problem is the fact you don't pair it with "style" lora. This is all-you-can-eat model, but needs you to choose "focus", which is accomplished with pony style lora

skill issue

You have to actually pair it with at least one style lora for it to produce good. It is intended to work that way

А какая там основа? Flux?

@Anton0695 Недообученная, брошенная создателем, кустарная Auraflow за версией 0.2, когда последней была 0.3. Мало того, что недоделка, так Астралох умудрился сделать её ЕЩЁ ХУЖЕ.

Не понимаю, почему просто не поменять основу. За прошедшее время это можно было сделать уже несколько раз.

@Anton0695 ему предлагали, но тот упёрся из вредности ("назло бабушке отморожу уши").

@Chertilo Это очень многое объясняет. Однако, очень жаль что так произошло.
Хотя нам, как пользователям, просто жаль, а вот автору - заслуженный удар по репутации. Всё-таки правило "Обещал - сделай" никто не отменял.

@Renesi база.

i think for that you would probably use either no skin color tag, or "light skin", and mature male in the negative? from what ive seen mature male tends to mean man in 30-40 range, with "tan" getting medium skin tone and "dark skin" or "dark-skinned male" for darker still.

>>> Illustrious XL/NoobAI XL

crazy how i got downvoted to oblivion for asking how to have a white male lol

@PsionCrystallis i tried everything but it doesnt really trigger, im using another model thats mixed with pony so maybe it takes more inspiration from that training data rather than ponys

@theyakuzi  aaaah, Im not sure then. it gets weird when remixes are involved, especially since not all remixed models are.... good remixes. For instance, porncraft, despite its name, was absolutely terrible at gay or furry porn. <_<; seemed like it was entirely stacked toward female focus and your standard vanilla couple. Basically, pay attention to a remixes goals and strengths for the objective you have in mind, because non-base models typically are heavily stacked in some way. some only are good at 2d or 3d in a certain style, they may specialize in furry or human art, they may be porny or SFW, focused on people or detailed landscapes. Once you know a model's strengths and weaknesses go from there.

Also, tagging systems vary depending on the model and the bases it uses (or the base of the remixes it uses). pony uses danbooru tags i believe, so check out danbooru's wiki to find out how to properly tag stuff for it. but if it's a remix of say, noobai or illustrious (which a lot of illustrious models actually have noobai mixed in), it might also use e621 tags as well, which is basically furry danbooru.

(and i'm not the kink police, we can't help our fetishes and attractions. Like, I don't even know if you're a white guy who screams at the sight of melanin, a black guy who likes fair skin, someone prompting to make a specific character... and none of that matters anyways. as long as you're not being a jerk to others what you find sexy are your own beeswax. If you're even prompting for naughtiness at all and not simply just making an ordinary picture, lmao.)

@odol I mean it has nothing to do w bbc i just want a white male , idk why people read so much into it. Weirdos

@theyakuzi i suspect a lot of it has to do with pent up helpless rage from the state of the world these days, especially the USA. Like, almost everyone is stressed to heck, countries are weaponizing bots to make ragebait comments and divide people so they dont pay attention to the ploys of the rich, and nutrition is down and pollution up. No one can afford mental or actual health care here, and news media of all sources blasts fearmongering and lies.

So i suspect people are cranky and easily triggered by anything due to thier understandably frazzled statw of mind, and internet screaming is thier poor mans therapy. Complain, yell, white knight, anything to feel better and somehow useful in a world that wont allow them to actually make the world a better place.

Unfortunately, such actions only further divide us, making us easy to control and manipulate. I include myself in this, as when i was a child and young adult i felt hopeful for the future and that i could be a part of that. Ive noticed the last couple decades has been a slow decline in my own patience and some emotional outbursts from the sheer frustration and helplessness ive been feeling. I suspect most feel the same, if not worse.

Defending minorities on the internet is good and all - i am one myself - but people wont change thier viewpoint from screaming, only harden thier beliefs. And thats even assuming there is an issue to begin with and not simply miscommunication or someone assuming too much.

@PsionCrystallis This assumes minorities need defending in the first place. If they all acted like you nobody would need to spring to their defense, because there would be no reason to dislike them beyond mere petty racism. The world is slowly going to hell, and people would only change if they were willing to listen to the other side, like you do (and understand that it's really the elite trying to divide us - but this is neither here nor there). But that's definitely not going to happen. Still, thanks for being a sane voice drowned out in a cacophony of idiocy.

@odol unfortunately, they do need defending. I know this, as a minority in several ways - petty racism is all the reason some people need to dislike you, because thier own lives are so miserable and they desire to feel large and in control so badly they will gladly tear anyone down for any reason to try and look better in comparison. I still remember as a kid how common it was to see yet another story about a man being tied and dragged behind a car for being gay.

That being said, it really is much worse in some ways now. Legal protection has gone up, but political wars and manipulation have amplified extremes on all sides to the point calm reasoning is drowned out. Just like any relationship though, screaming gets you nowhere. Open communication is important, because all involved parties can have valid points or concerns that you never thought about. One of my best friends had a lot of repressed anger because he was white and constantly bombarded online about how he was trashed simply because of his skin color and was entitled and should die for it. Thats just racism hiding under the guise of justice, whuch is a favorite trick of bigotry. Helping reassure him that he was NOT garbage for being born was the confirmation he needed to start loving himself, and from there, others as well.

Im not sure what the answer to fixing all this poisonous hatred is, but all i can think of is how both hatred and love appear to be contagious. Thus, i try and find what little bits of hatred i can find thats managable, isolate it, and counter it with a greater amount of compassion. Nurture enough flames of love and eventually you get a sun.

Im not going to lie though, taking in a person's hatred, bigotry, rage... its work. Like singing a song and having someone shout at you. You have to keep singing until they exhaust themselves and join in. You can feel depleted afterwards and need quiet and peace to recharge.

But ive never regretted the attempt. Its led to wonderful people and new experiences.

because v7 v8 v9?

@ashadowgirl i take v8 over v6

just a 'score_...'

It was a joke

fever dream, definitely

StabilityAI told civitai to remove Turbo

@qek wait fr? why?

That fictional is shi*.Just a fkn scam.

It's for 12 years old kids lmao

Ive not really used Pony in general, but I can def agree, Im not sure what happen between V6 to V7, but its quite wild when I looked at these 2 models.

@ThatSoKitten v7 is based on auraflow, despite people warning him.

That's why it's shit.

Turbo is distilled, also, the Turbo version with no DPO has been removed

For 'TURBO' model use Euler A with 15-20 steps and CFG 1.5-2

Why?

I didn't really think of that

I think so. It should be

Right? I tought I am crazy, because I used NetaYume Lumina at first and only dropped it because I didn't find any LoRA that I cared about, but pony is just ass and it sucks at making hands. Lumina does hands perfectly almost every time.

According to the AI, this checkpoint trainer is the one that works best with perspectives and camera angles, better than others of higher quality. Even though its quality may sometimes not be what one expects, requests for isometric perspective from various angles—top-down, frontal, lateral, etc.—will always come out well (regardless of whether the overall result is what you expected). That makes it useful for redrawing later, adding details afterward, or merging with images you wanted but whose proportions didn’t match. That said, I still haven’t tried it yet.

Pony came out in the beginning of 2024 mate. It was revolutionary back then.

If you want to know why it go so much attention, try the SD 1.5 checkpoints that came before it in 2023.

while it may not be as advanced as current models nowadays, i personally think pony can do quite a lot, at least in terms of character art.

When it was released, it was the only 2D model that could generate halfway-decent NSFW.

Oh, only like 5 months ago?
https://civitai.com/models/1901521/pony-v7-base

Dont bother using it, its dogsht compared to this one

@TheRedRhino And this one is dogshit compared to most IL and NAI.

it doesn't exist

PonyV6 = SDXL
so you can use any SDXL workflow.

@Silvicultor thank you

Better off with IL models now since (SD/SDXL/Pony) is outdated check out: WAI-illustrious-SDXL

You downloaded the pony base model

go with pony realism, or better big love

Also what I'm wondering. I was generating with it just fine yesterday, now it says I need a subscription to use a "private model."

@Lakleg subscription doesnt matter. it locked regardless

This is a technical glitch, I had the same problem with another model. I emailed the model's authors, and only they can fix this, so if they didn't do it on purpose, and this really is another CivitAI glitch, they should fix it soon. We’ll just have to wait.

The model might be updated. I saw the link/model number changed, so just update the download URL/link of the model in your interface and it should be fine.

@strashnaadresa875 There is Pony Diffusion XL V6 and there is Pony Diffusion V7, but these are completely different models; V6 hasn't been updated in a long time.

This is a technical glitch, I had the same problem with another model. I emailed the model's authors, and only they can fix this, so if they didn't do it on purpose, and this really is another CivitAI glitch, they should fix it soon. We’ll just have to wait.

The model might be updated. I saw the link/model number changed, so just update the download URL/link of the model in your interface and it should be fine.

This is a technical glitch, I had the same problem with another model. I emailed the model's authors, and only they can fix this, so if they didn't do it on purpose, and this really is another CivitAI glitch, they should fix it soon. We’ll just have to wait.

The model might be updated. I saw the link/model number changed, so just update the download URL/link of the model in your interface and it should be fine.

@strashnaadresa875 There is Pony Diffusion XL V6 and there is Pony Diffusion V7, but these are completely different models; V6 hasn't been updated in a long time.

This is a technical glitch, I had the same problem with another model. I emailed the model's authors, and only they can fix this, so if they didn't do it on purpose, and this really is another CivitAI glitch, they should fix it soon. We’ll just have to wait.

@eXP_dot_Net Are you really sure that they're gonna fix this model?

@Chris02 I can't speak for them. But on other AI services with generators (TensorArt, SeaArt, PixAI) this model works fine, without any crashes/problems/blocks. I had a similar problem with another model, I wrote to the author and within a few days he fixed everything. I wrote an email (it's in the model description and on their website) to the authors of this model. I hope they'll notice and fix it. They've distributed it not only to CivitAI, so we're waiting for it.

@Chris02 I certainly hope so. Personally, I haven't used Pony for quite a while now; these days, I rely more on "Nova" IL or SDXL. But in theory they should fix this, especially since they state that this product is under a free license, and only derivatives of this product (that is, images generated using this model) can be used for commercial purposes. The fact that you have to pay (buy a subscription) to use Checkpoint itself is already a commercial use of the model itself, which is directly prohibited by the license of this model. Moreover, there was a comment here that even with a subscription the model cannot be used now.

@eXP_dot_Net Ok. I'm pretty sure they'll fix it.

The model might be updated. I saw the link/model number changed, so just update the download URL/link of the model in your interface and it should be fine.

The model might be updated. I saw the link/model number changed, so just update the download URL/link of the model in your interface and it should be fine.

@strashnaadresa875 There is Pony Diffusion XL V6 and there is Pony Diffusion V7, but these are completely different models; V6 hasn't been updated in a long time.

thanks for providing an actual answer. so who fixes this, Civit or the new/old owner? and any idea how long that would take?

@waifu4u2 Has been fixed (by civitai devs)! It was probably the commit mentioned above.

it was a bug in civitai's side that caused some models and such to become unpublished until they could fix it so its probably from that