you like my content, support me at:
Patreon in here:
https://linktr.ee/Duc_Haiten
I created my discord server. I'm still quite new to discord server management, free to give opinions.
special note: Using the model is very simple, like the scores, but also not like the scores. The idea behind it is training based on various artistic styles of different artists and some unique AI models, arranging them into separate keywords (perhaps I should have made them numerical codes like score_1234 in Pony base, such as style_123—maybe next time I’ll do that). Each keyword corresponds to a different style, representing an artist or a uniquely special AI style. Users can call up just one keyword to bring out that style, but that’s not the main intention. What I aim for are unique style combinations. You can call up multiple styles at once, for example: A, B, C, or B, C, A, or C, A, B.
Any change in the keywords or the order of the keywords will result in different outcomes. You can even use parentheses like (A) or (A:1.2) to increase or decrease the weight of a keyword. This combination and arrangement will turn the initial 55 styles into hundreds or thousands of different styles, and as long as you don’t reveal your style combinations and keywords, it will be a one-of-a-kind creation unique to you. That was my original intention when creating this project.
Read more here:
https://civarchive.com/articles/8523
Because the image generator of civitai and automatic1111 are different, on civitai you need to set the number of steps about 10-15 steps higher than when using automatic1111.
Update on October 16, 2024:
Pony No Score v7, this will be a groundbreaking version that unlocks enormous potential for pony, marking the largest update to date. V7 integrates 55 styles into one model, each with separate trigger keywords, similar to how Pony previously added scores to the model. The goal is to create a model capable of producing a wide variety of artistic styles, rather than being limited to just a few that are difficult to control. Users can apply individual styles or combine different styles to create their own completely unique artistic vision. Once you hide your prompt and data, no one will know which style keywords you combined to craft such a distinct look—it will be uniquely yours.
I will write a detailed guide on how to use the model along with a list of styles here, so everyone can gain a clearer understanding of what I’m aiming for.
https://civarchive.com/articles/8105
Update on September 8, 2024:
Finally caught up with pony base v6 LOL. This is probably the last update, aside from bug fixes. I feel like I've reached the limit with pony v6, and adding anything more might cause issues elsewhere.
Version 6.0 includes some anime data, better adjustments for NSFW, further tweaks to colors and lighting effects, and additional improvements in natural language.
And the most important thing is earning money. Don’t be fooled by how popular this model seems. In reality, I work 15-16 hours a day, without taking a break, just to earn enough to reinvest in research and model training, with barely enough left to live on. So now I want to see how much I can make with this early access feature on my most popular model.
Update on August 8, 2024:
v5.1+v5.0+v4.0=v5.2
Damn I forgot to turn on early access
Version v5.2 works quite well both with and without scores
Update on August 4, 2024:
Version 5.1, a complete overhaul of the text encoder and adjustments to the U-net. Efforts have been made to minimize minor bugs from versions 5 and 4. Once again, the default importance level of the 3D style has been downgraded. I do not want 3D to be added to images if it is not explicitly mentioned in the prompt. Of course, if reducing the 3D style too much makes many people feel dissatisfied, I will adjust it again. So please let me know your thoughts on this adjusted version.
The parameter settings I have tested:
- Sampling method: Euler a
- Schedule type: SGM Uniform
- Sampling steps: 30-36
- CFG Scale: 8-9
Update on July 23, 2024:
Welcome to pony no score v5. I tried to save it until pony base v6.9 was released, but the waiting period was too long, so I couldn't wait any longer and got to work immediately. I need it to have more soul, more artistic depth, and to convey emotions in each created image. I am fed up with the dryness of AI-generated art, which only looks beautiful but feels empty, unlike something truly nurtured from the creator's soul. I don't know if you understand what I mean, but that is what I am trying to achieve. Perhaps the most noticeable change for you is that the 3D style will be minimized, and the 2D drawing style will undergo slight modifications, with colors feeling more cinematic, and the lighting and shadows will be better and more realistic.
Some other updates: improved text encoder, more shooting angles, more art and 2D data to replace Nijijourney data
Recommended settings:
Sampling method: DPM++ SDE
Schedule type: SGM Uniform (this one is a must)
Sampling steps: 30-36
CFG Scale: 9
update May 27, 2024:
This will be the final version until pony base v7 comes out. In addition, the alpha 1 version of the pony-real model has been released, everyone can check it out, but don't expect much from the alpha.
Thank you very much and apologies to Deathless, who helped me find and process around 8,000-10,000 image data for the pony-no-score version 4.0, whom I forgot to mention. So much happened in a short time that it gave me a headache. Fortunately, he reminded me, at least he didn't curse me out harshly LOL.
(By the way, currently I'm doing everything alone again, if anyone is interested and willing to help prepare data for training my model, please feel free to contact me.)
update May 22, 2024:
Final beta version of v4, the next update will be its official version. When the official version of pony no score v4 is released, it will also be the day pony-realistic alpha will be uploaded
updated May 14, 2024:
Version 3.5 for Pony-no_score.
- Edit the text encoder
- Update more 3D image data
- added more realistic skin models for 3d
- reduce unnecessary details that are not in the prompt (creativity is a bit reduced but accuracy will be higher)
- Enhance solid contours
This is a separate version from v4 in training, so I need the most objective reviews.
Update May 7, 2024:
I must say that I am really surprised and happy to see everyone's warm reception for the DucHaiten-Pony-XL model (no-score). Each new version gets more attention than the previous version, it seems I'm still on the right track.
To express my sincere thanks to everyone, I will post the v4 beta version first, everyone can try it out and give advice and suggestions for it. Note that because it is only a beta version, there will be many changes in the future.
Update May 3, 2024:
After hearing comments about v2 seeming to lose the diversity and balance of styles compared to v1, this time I tried to adjust that problem. In addition, a lot of new data has been added in v3. They are arranged to intersperse with Pony's previous data, with the purpose of supplementing and creating connections with each other. I do so with the desire to create a more complete ecosystem for ponies. But because I'm the only one doing it all on my own with such a small budget, I can only try to limit my abilities as much as possible. PS. Colors are more vivid and brighter.
updated April 28, 2024:
The upgraded version 2.0 will have many major changes, first is adding more handsome men data, next is improving weak natural language capabilities, and finally adding more artistic feeling to the model. There are also some smaller adjustments like details and lighting
A model created from the original pony model and combined with my various models. I tried to keep almost all of pony's text encoder and only change the image data, so I recommend you learn more about pony if you don't know how to use "tag score" properly. When using this model, you will feel that the image is less random and don't has too many unnecessary details like the pony, and will feel smoother and cleaner, Eliminates the possibility that some prompts will produce harsh images or burnt colors, the face is made clearer, the hands, feet, and pose are adjusted for fewer errors. But if you gain one, you lose the other. It can only be said that each person has its own style, depending on your preferences.
If you want to use my models to create unlimited images at a small cost without having to install anything, mage.space will be a good choice, i have some models that are exclusively available on there.
Description
FAQ
Comments (202)
I recommend version 5.0, it works with some loras that no other checkpoints seem to work with (like the Toshiki Yui lora)
55 style in v7
Arcane
ttguweiz
carmesjocelin
fiz-rot
anninosj
samdoesarts
elliemaplefox
sei5osuzuki
BM94199
yaoyao794
sea_Pall
Vergilhoo
hamukukka
omao_51061954
letsfinalanswer
106858
sungmoomoo
hakwansoo36
wonbin_lee
hex_jjung
Ghibli
niji3d
Hoovesart
cutesexyrobutts
afrobull
unfairr
Khyle
shexyo
Sinozick
ironlilyart
mossacannibalis
gabrielpicolo
Rossdraws
Kuvshinov_Ilya
RDJlock
wlopwangling
beitemian
strauzek
RyanReos
hdfurrybabes
Apone3D
wildeerstudio
AkiyamaRyo
fugtrup
crazytown
748cm
ArawAraw
stevencarson
rescraft
aromasensei
ponyvid
ponycos
ghost236
7200ss
lizchiefffff
Are you planning on having a showcase for each style, or at least a description ?
@Funky007 I think it's more fun for everyone to try little by little and discover their own way of playing
@Funky007 Just generate X/Y/Z plots with different combinations. I believe for the community it will be easier to perform random walks... "Let's map together the Earth-like planets scattered across the galaxy..." (E:D reference ;)
@DucHaiten I am curious what if I merge the CLIP blocks from your model with UNet blocks from some more realistic models... Don't you mind I will make yet another merge?
@homoludens Please mention my name and my model, help me advertise this model, it's all ok if you do
@DucHaiten OK.. will do the same as with previous merge in CinEro Pony series.
@homoludens Can I do that on Civitai ? (Asking seriously, not trying to be a smartass)
@Funky007 I did not see any instruments like X/Y/Z Plot for A1111 WebUI. But I am sure it is possible to run SDXL/Pony model on quite low grade GPU like RTX3050 8Gb
HAHAHA... LOL! I've used your model (v7.0) as a LoRa inside-out.
I applied my GaalGai realistic LoRa made of photos from mountain hike and asked your model to stylize my LoRa instead of vice versa... Such a surprise :D
Will post images soon...
Is it available on site?
Its in early access. So in 2 weeks it will be. Or if the 500.000 buzz goal is met
@Meower2024 if people are willing to spend buzz then the fastest it will take is probably only 5 days, but i guess it may not be that fast
@DucHaiten These models are def worth the shared 500.000 also. Its one of the most used models. Must have taken ages to make.
@DucHaiten pls can u put the prompt on the showcase picture
@Mona_ok I was going to let everyone figure it out for themselves, but it seems like this new model is too new for most people, so I'll turn on the prompt.
@DucHaiten thanks
A1111 Model Toolkit reports that v7.0 is broken. Could you please verify integrity?
I tried to make quick merges in different modes. Model gets broken. And toolkit sees only VAE module.
~~~~
Statistics
Total keys: 2514 (6.46 GB), Unknown keys: 2266 (6.31 GB).
Architecture
VAE-v1-BROKEN
VAE-v1
VAE-v1-SD
Rejected
UNET-v1-SD: Missing required keys (446 of 686)
model.diffusion_model.output_blocks.9.1.norm.weight [320]model.diffusion_model.output_blocks.7.0.emb_layers.1.weight [640,1280]model.diffusion_model.input_blocks.1.1.proj_out.weight [320,320,1,1]model.diffusion_model.output_blocks.9.0.out_layers.3.bias [320]model.diffusion_model.output_blocks.6.1.transformer_blocks.0.ff.net.0.proj.weight [5120,640]...
UNET-v1-Inpainting: Missing required keys (447 of ...
UNET-XL-Refiner: Missing required keys (1219 of 1220)
...
CLIP-XL-AUX-SD: Missing required keys (1 of 197)
conditioner.embedders.0.transformer.text_model.embeddings.position_ids [1,77]
...
SD-XL: Missing required classes
CLIP-XL-AUX
This is how SDXL architecture should look like:
Architecture
SD-XL
* VAE-v1
- VAE-v1-SD
* UNET-XL
- UNET-XL-SD
* CLIP-XL-AUX
- CLIP-XL-AUX-SD
* CLIP-XL
- CLIP-XL-SD
Really nice but been getting alot of watermarks/signatures
wlopwangling seems to have quite alot of watermarks coming with it from what I've seen so far
@Hati_Hrodvitnisson Adding "watermark, logo, author name, web-address" to negatives does not help?
@homoludens Nah not for me, I even added weight to it but still got them. but it seems to be more common for some styles than others
@Hati_Hrodvitnisson Also, try to reduce the weight of trigger words. For me triggers with weight 1.0 or above produce instabilities and for some sampler always produce the broken picture like fractal.
See my pictures posted here for particular weights which does give good results.
Very cool
Now that I've used it some more I've come to realise that it's first of all really fun to use since theres a large variety of styles without the need to download extra loras or anything, but what bugs me is that they can be unstable, within the same batch I can sometimes get 3 different styles (batch of 6 images). I really hope this will be solved somehow in a future version because this is already incredibly fun to use and would be even more so with more stability.
Can you talk to me more about your problem? I want to know clearly how you used the trigger and prompt keywords and it randomly appears different styles, at least I still Haven't had that problem yet
@DucHaiten i had similar behaviour sometimes. I usually make something like 20 images in a row with wildcards and fixed styles like "score_9, score_8_up, score_7_up, source_anime,(style1:1.2),(style2:0.25), (style3:1.05), " in ComfyUI. Some pictures came out rather different .. maybe because of different "camera angles" or poses
@Sithjid You can try putting the style section at the top of the prompt, so it will have more priority
And also play with the weight of Pony Scores. I see pony scores interfere with styles noticeably.
Issues with stability looks fixable to me (via scores weight).
@homoludens or simply increase the weight of the style
i already have the styles at the beginning after the scores. But the biggest effects seems to me .. if i change the order of the styles. Like, at the moment im trying four styles at once. And when i just change the order in between of them .. the image style may change "a lot". the impact is bigger than weighting
@Sithjid that's normal, that's why I said it from the beginning, it's a feature not a bug, imagine it like a series of passwords, the position before and after will also change that password to be different from the original password, only then will people not know what you're using
@DucHaiten Yeah ofcourse, though I don't have any images saved from that batch since it turned out weird, but I was using a couple of the different styles and made a kinda neat anime like image but using an anime character (I think that's part of where the issue comes in). I was making an image of Chichi from Dragon ball. and it was a pretty varied style in batch
@DucHaiten Oh and just to be clear, I don't mean any of my critique as mean or hostile, this is by far the most entertaining model I've used and I simply hope for it to get even better if possible
@DucHaiten i didn't mean to call it a bug. It's quite interesting to use :-)
gg
v6 is my favorite had so much fun got it early with buzz and the results are pretty nice with lots of ways to control style
but v7 i don't know about this one. Seems very off and not appealing at all, the results other post are very janky and as i heard and see reviews yeah people are facing lots of inconsistency.
v7 is more about combining styles the way you want
Agreed! Maybe for my use (semirealistic) doenst fit anymore. Or, a have to find another keywords.
@123123444 When this model is out of early access, I will make a specific article explaining the capabilities of this model
@DucHaiten yes we need an extensive guide for this one thank you for your hard work
Congratulations on releasing v7. It's a great all-around model and I'm still playing with different artist styles. Good prompt adherence, aesthetics, everything. Nice work!
thank you, but it seems some people don't like it, they seem to like the old model instead of having to get used to something completely new, it seems like too many trigger keywords make them confused
@DucHaiten I see this more as a new base model like Pony, rather than a v7 of your previous models. Maybe when Early Access is up you can rebrand it/ launch it as a new model with some context that makes it intentional they have to create or find their own style?
@prompt_bit_sorcerer I don't want to because civitai only allows me to turn on two generators at the same time.
@DucHaiten Ah. Maybe a version rename then? Some way of setting new expectations.
I second the congratulations. It's been a while since I had that much fun with a pony model. True, it takes a bit of learning and experimenting, but this seems to open endless possibilities. Great job !
now. how can I use V5.2 to create??
Mage.space and locally
10550 Buzz transferred as a split donation from XenEro v1.2
Thanks 👍👌
A Deep Guide to DucHaiten-Pony_no-score_v7.0
https://civitai.com/articles/8523
I was very excited to see your v7 model released! Is there any way to get it available for Civitai's generator?
is on now
the new 7.0 version is not good for me😢..please make version 6.0 usable
please read my article
Yeah, I would like 6.0 back too.
@DucHaiten is it possible to leave v6 as it is, it was almost perfect, and then perhaps upload v7 as an entirely different model, as v7 really is a whole new initiative(a good one tho)
kinda feel like the whole 40+ artist model could have been a different model altogether?
buttttt.....ngl it's kinda fun experimenting
@evandesilva please read my article
It feels...like it got worse instead of better. Could we also get 6.0 to use again?
I would also like 6.0 back, I had just got a really good flow going with that one.
read the instructions here, of course i want all previous and future versions to work too, but civitai lacks the GPU to do that
@DucHaiten I have. This version is heavily focused on styles and personalization which on itself is great. But I feel like it would have been better to upload that as a seperate model because of the focus on style. Please don't get me wrong, I absolutely love what you've done with training models and 5.2 is what made me use CivitAi. That's why this is a bit of a bummer as it feels like I cannot recreate the same quality of 6.0 with the same prompt as now I'd have to find the right style/combination of styles. But maybe I haven't generated enough.
@Missing0000 yesss, same, I have some projects in the pipeline that would perform the best with v6, I also feel that since v7 is an entirely new initiative(and a good one), it would serve the community better if it was uploaded as a separate model altogether... @DucHaiten
@evandesilva I did it just to have a generator for it
@DucHaiten Is there supposed to be a combination that has the previous 6.0 look or something?
@CryptidMan I'm managing to get closer to my v6 quality, but with some experimentation I've found that with v7 I can actually get even better, but I have been spending hours on figuring out all of this and mixing up different styles, prompts etc. I am assuming it's the creator's intention to make us experiment and find our own style here... idk.... it's been hours of experimenting for me, i'm seeing same great results with the ability to add new elements
@evandesilva Exactly
@evandesilva Instead of a simple model, I think people are gradually being able to make their own mark, the feeling of tinkering and creating results will make people more excited.
@DucHaiten it's time to level up, thank u for thinking ahead and giving us this opportunity
I read the article. I've also used both 5.2 and 6.0 quite often, as can be seen from my gallery. They were temperamental models but with lots of potential. Now, anybody that will try to remix my existing images will fail. I don't know if any of the dozen ongoing projects that weren't fully tweaked will be able to continue from the existing state, or if I should start from scratch
Fyi, what I said doesn't mean that the new model isn't a step in the right direction. Time will tell that part. What I mean is that the choice to abruptly and irreversibly cut the other versions is imho very shortsighted. That said, thank you for your efforts.
@revenge623 all the old version is still there, it's not gone, only the generator is gone because civitai lacks GPU
@DucHaiten I understand the situation. I create all my images on the site, I don't have a local deploy and the fact that the older versions are available for download does not help me. I'm sure you thought that this was the right step, at the end of the day these are your models, and I have lots of love for your work ... that said, proper communication and transition was missing and we as creators are shouldering the consequences. Stopping something that people have learned how to use and demand that they learn something new overnight instead (whether more revolutionary or not) is misguided.
@revenge623 It's because civitai lacks a GPU, I can't let the new v7 come out without a generator
@DucHaiten I don't want to sound like complaining because I know that trade-offs are made when real life is involved. Part of what I'm saying can be used as future feedback or as giving a different point of view. How you use it is up to you...
I fully understand the reasons and the big picture, and the two articles (yes I read the Oct 16 announcement as well) are informative, although I don't think I read anywhere that by Oct 31 is a hard cut-off date for the previous version.
Even without that, there's no clear guidelines how one can use the existing model to achieve "at least" the same things that they were able to do with previous models. I'll take for example Sinozick since it's one of your included keywords. Previously, I'd use your checkpoint with Sinozick LORA (let's say strength 0.75) to get a certain style, colors, etc. Does this mean that now I should use both the Sinozick Lora and also the Sinozick keyword. I've tried using one or the other or both, all my attempts so far have fallen short of getting anywhere close to the previous quality.
Personally, if I do a blind compare of just the example images that you've posted for 7.0 with those of 6.0 and 5.2, I'd never consider the 7.0 as a superior model.
Also if you think that the current model has any weakness or things to avoid, that would also be useful information to have (going based on what you mentioned in the announcement article)
Much love....
The mixed bag reaction is a given based on the purpose attempted by v7, I am starting to enjoy the experimenting process more. Thank you <3
I do have to point out that the Hand quality has gotten especially bad though, any advice on how to minimize this? thanks
V7 is a failed experiment. Almost impossible to get good results out of it.
I see people still creating great results, see how they do it, and read my beginner's guide to v7
the examples seen below seem to suggest otherwise
listen guys before using please read my explanation about v7 and its exact usage instructions.
Please tell us how to prompt to get your original style like in 6.0
This version is good at prompt understanding but your old styles are a lot better!
Yeah I read the articles. They're not helpful. Keeping it secret which keywords do what things is, in fact, the opposite of helpful. I liked the way things looked in 6.0 and I don't enjoy the idea of having to do a bunch of random experimenting with arbitrary keywords to get back to a place I was already happy.
I personally like the idea of having trigger words for different art styles but the original style gave this model it's identity which seems to be lost with 7.0. It would've been best if new style did not have any influence unless the trigger word is used. imo you should add [experimental] to the name and fix the issues with styles in next update. The model is good but needs some work.
Anyways thanks for your work, it's always great so don't mind the complaints, it's just criticism everyone faces once in a while.
Thank you for adding the links. I'm begining testing all the combinations with the previous prompts. This is will be fun
I've been going thru the trigger words. very stable when paired correctly. I search for an example of the trigger words for reference and I get consistant results.
I also wish civitai would give me more generators
@DucHaiten Yes, that would be the ideal. Then we could keep v5/6 around in the generator and still have this for those who want it. But i have to agree with ImmortallDeaad, this version has to a large extent lost the distinctiveness that made the earlier versions my go-to Pony. There are already plenty of loras for specific artist styles (thank you Vixon) that mostly play nicely together if you want to blend them, so building them in at the cost of the "base" style feels like a retrograde. This version can still produce flashes of brilliance - but it lacks the consistency one needs when committing buzz in the onsite generator. Is my two cents worth.
@SimianHose @ImmortallDeaad I agree with both
I absolutely adored 5.2 and 6.0. Please DucHaiten. If it's alright, could you please allow either or both of those to still be used?
if you make a form and get everyone to sign it and send it to civitai admin then maybe it will work
guys, i don't have permission to open more generators, i have two, one for pony real, one for no score, do you guys want me to move the one on pony real to the one on no score v6?
Yes.
I would personally prefer having access to"no score v6" over "pony real".
Thank you for all you hard work and for being receptive to feedback.
@MuscleGong that's because this is a no score post, if i really do that i will fuck up
@ImmortallDeaad I texted MAX asking for another generator, but he hasn't responded yet
yeah. I think that's a good idea. But i don't want it if it's become bad for you.
Everyone misses v6.
yess I do love v7 as well but I thought that v6 was perfect, I think the community would love having access to v6 and v7 separately
@DucHaiten any news?
I really like v7. The outcome of combining styles can be very gorgeous. It takes a lot of experimenting, which is quite nice actually. But i think, if you're not able to run it locally, it can be quite frustrating. If you're used to style loras, just don't use them in this model and try combining trigger words.
here are some i used so far, maybe it's useful for someone:
(Sinozick:0.6), (Arcane:1.1)
(Arcane:1.2),(Sinozick:0.25), ttguweiz,
(Arcane:1.2),(Sinozick:0.25), (sea_Pall:1.05),
yaoyao794, samdoesarts, ttguweiz,
Sinozick, sea_Pall, arcane, samdoesarts, ttguweiz
Sinozick, sea_Pall, arcane, carmesjocelin, ttguweiz
wildeerstudio, arcane, ArawAraw, 748cm
748cm, ArawAraw, wildeerstudio, carmesjocelin
I like 5.2 and 6.0 very much too
People hating on v7 don't understand how to use it. Make sure you take a look at the guide and experiment with different trigger words to get a style that you like.
Probably just prefer the 6.0 look.
I've already deleted v6 after using v7 for 2 days. It's a bit harder to use but the fact that you literally have kitchen with every possible tool for cooking is amazing. I don't have to download LORAs for everything I do is simply awesome.
Also v6 has it's own identity, some might prefer it over having multiple styles. I personally think having a trigger for v6 style would be amazing.
Yeah, yeah people are stupid...for a user, things need to be simplified, not complicated. We already have a thousand other settings to tweak before the 80 different styles and combinations. I'm glad you're enjoying the model.
I have read the articles on styles and experimented for several hours with v7. IMO this model is too difficult to use to be useful. This is proved by the fact that most of the images posted for this model are just using style Loras. There are hardly any images using the in-built styles.
Typically, I pick one or two styles that I like and then experiment with prompts to get the content that I want. This model doesn't have enough documentation or sample images to help select a style that I want. With style Loras, we can easily find them on Civitai, and get down to business. It's just impossible to find the style I want with the thousands (millions?) of style combinations that are possible with these 55 styles.
v6 was a very versatile model that had its own style, and worked well with a large number of style Loras that literally have thousands of sample images. v7 is just very frustrating to use.
V6 back
I think V6 was simply perfect, so thanks :-)
Thank you for listening to the feedback. I also appreciate you trying new things, I hope people running it locally will make good use of V7
Wooo, thank you so much.
Thank you so so much for bringing v6 back. I know you worked extremely hard on v7 but I don't think it's quite there yet. This is by far my absolute favourite model and I can't thank you enough for all you've done.
That's a shame with v7 no longer on the generator I was having fun generating different styles, no other model can do that on its own, at least that I've come across. I hope you don't stray away from the idea of adding styles to your model in the future. I think its very innovative and sets it apart from other models. I think people just didn't give it a fair shot and test it out enough. But totally respect your decision.
I still think it's an interesting direction, it's just too picky about users, but in the future, people will gradually get used to it and be more open to my other future models. Maybe the sd3.5, or when the newly released pony base
@DucHaiten Are you working on a sd3.5 model?
@Sithjid it needs to have a colab training version, i don't want to have to train locally
Absolutely loved V7! I mostly make images for myself, so I don't tend to post them, but the results were so good. I loved how the included styles never became too overpowering. It's a lot of fun! Hope the people complaining don't discourage you, because it was really, really awesome.
🥺I've just started to get used to v7. It was a bit more complicated than I had imagined, but I enjoyed the process. It's a lot of fun! Thank you!🤗
Brother, why do you listen to children and remove the v7 ? It was really great and versatile 🦆
i didnt remove it, just switched the generator to v6 because there were too many calls for it that i couldnt ignore it anymore. i heard civitai will have a public vote in the future to decide which model will have the generator, by then i wont have to worry about this. I have had enough
@DucHaiten Oh, good 👍
6.0 IS BACK thank you 7.0 is unnecessary complicated 7.0 just should be separate thing
You can adjust its strength like loras, in fact every keyword in the prompt can adjust its strength, that is not loras' privilege
@DucHaiten got it
@DucHaiten will u bring back v7 as a different model?
@evandesilva Does it make sense if is not have a generator?
@DucHaiten hmm I do think that your idea and intention for V7 is valid and if you have the time and energy you should bring back v7 but in a different generator. so that the community will continue to have access to it's beloved v6 which is almost perfect. Best of both worlds...if possible
@evandesilva Civitai won't be able to provide more generators for a long time to come, so if the user appeal is strong enough, maybe Civitai will take a generator from another creator and put it in v7. so this is on the user and Civitai, i only have two generators, the other one is on the pony real model which has more users than v6, so i cant take from there and give to v7
@DucHaiten understood bro, I think the community understands that you are doing the best with the given constraints, thank you for your work
I have had a very good experience with v7 but you removed it 😭😭
many people feel it is too difficult to use compared to v6, you can read all the comments I received
@DucHaiten yeah i know. I think it's because i used some LoRA so the results i got were more beautiful than v6 🥺
Btw mình cũng là người Việt 🤗
@Mia82 thấy tên với ảnh là đoán đc rồi
older versions delivered far better results
@DucHaiten 7.0 works well in Draw Things so far. I haven’t tried the new style prompting, but I was very happy with 6 and so far very happy with 7!
Not sure what happened to the styles list. Really don't feel like clicking a thousand links to get to the info I need to use this model. Maybe Im blind, or just rushing into judgment. Happens often to me.
Edit: I see the link now....styles list should be the first thing on this page. Not a find the link puzzle.
Arcane
ttguweiz
carmesjocelin
fiz-rot
anninosj
samdoesarts
elliemaplefox
sei5osuzuki
BM94199
yaoyao794
sea_Pall
Vergilhoo
hamukukka
omao_51061954
letsfinalanswer
106858
sungmoomoo
hakwansoo36
wonbin_lee
hex_jjung
Ghibli
niji3d
Hoovesart
cutesexyrobutts
afrobull
unfairr
Khyle
shexyo
Sinozick
ironlilyart
mossacannibalis
gabrielpicolo
Rossdraws
Kuvshinov_Ilya
RDJlock
wlopwangling
beitemian
strauzek
RyanReos
hdfurrybabes
Apone3D
wildeerstudio
AkiyamaRyo
fugtrup
crazytown
748cm
ArawAraw
stevencarson
rescraft
aromasensei
ponyvid
ponycos
ghost236
7200ss
lizchiefffff
I put it everywhere, in the comments, under the description, on my profile, but still no one sees it
@DucHaiten lol it's the first thing i saw, some people will complain about anything
This might be a horrible idea and maybe not viable, but since there's only one slot available, personally I'd be interested if there was a rotation of the two (three?) models...let's say every two weeks... That way, the people that want to experiment with one of the models (6.0 or 7.0) would be able to do so without fully committing to only that particular model ... since they can go back to the old model in a couple of weeks. Hopefully that way, the users would get more gradually accustomed to 7.0 before giving a farewell to 6.0
Bro, don't listen to anybody. Version 7 is the top! I found and created my unique style! Keep developing in this direction! All the best!
I believe 5.2 is more popular, estimate by the downloads, if 5.2 back to create will be right!!
Hey, just a question, the preferred setting for the v7 checkpoint is Euler A, SGM Uniform scheduler with 30-36 steps and CFG 8-9?
I read this in the description but I'm not sure it was referred to the last update of the v7.
can we get inpaint model for v7?
How is this differnt from using style loras?
lora can't add too much, besides lora is not good like training into model. you can use style and add other lora
I have had so much fun using this checkpoint. It's like a giant puzzle you unwrap. I even used an LLM to predict what the keywords might be for my experiment. I listed the keywords at on the my posts. Great job!
love the baked in styles
This model is a perfection, there are so many styles to try and you can mix them, its so good, i finally can generate unique images, thank you
Why isn't version 7 of "DucHaiten-Pony-XL (no-score)" usable with Civitai's Generator?
because there is no gpu for me
@DucHaiten So it's their decision, not yours. And why can't they activate version 7 instead of 6 ?
@Funky007 v6 is the decision of many users, they feel v7 is too different from the way they used it before, so they prefer to go back to v6. I'm thinking of creating a new base model from v7's idea
@DucHaiten That's too bad. I don't have a computer capable of handling image generation locally, and version 6 is the one I hate the most. It's buggy and tends to output gibberish as soon as too many LORAs are used with it. I'm not blaming you; I'm just disappointed to be stuck with this version instead of versions 3 or 5.2…
@DucHaiten Rename v7 as "DucHaiten-CreativePony" !
the problem is not only because of the name, but also civitai can not give me more gpu
Is there a Google sheet with the art style name and their examples ?
este modelo de verdad sirve para anime tambien? porque estuve creadno unas imagenes y no es que salieran como pense
Amazing style!
This is amazing!!! Could you make an Illustrious version DucHaiten in the future?
Dear friends, please share some pony models that are good at generating realistic Asian women. I've been looking for a long time and haven't found any. Thank you
You can try https://civitai.com/models/567471/pony-asian-realism-alpha
or you could use something like https://civitai.com/models/1046441/yourasiangf28d with a realistic-style lora.
Or alternatively you can try using any realistic model you like and slap any of the Asian women loras on CivitAI on top of it
I'e posted one image, made months ago, of a girl with a sport car:
https://civitai.com/posts/11835371
The image was then OUTPAINTED to complete the missing side of the car:
https://civitai.com/posts/12027025
Yesterday I got an animation from Kling, I will post tomorrow because is quite good! : )
Here we go!
Cool work, keep it up!
@Gavare Thank you! But now I noticed, for some reason, the link shows a sort of small version of the video.
Please, check this one and tell me if is better!
https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a689f263-c2c2-4f1e-80d8-a2ac6b29e97f/transcode=true,original=true,quality=90/53971965.webm
@settima_ai It's same but with better quality, so yeah, better)))
V7 is VAE baked?
Almost all current models are required to have vae when training
This is the best..fir NSFW images
Are you planning to make an illustrious version? I love this model, I think it would be really interesting!
try my DHXL model
Why this model isn't available now?
I really like the idea here! Would you be willing to share your dataset so I can apply the same thing to animagine XL 4.0 in fine-tuning?
@DucHaiten what do you think?
For a long while this has been my favorite model. Pitty it's down :( Can't craft my hot demons without it.
I love this model. I've spent way too much time trying out different style combinations and it never gets old
This was my fav checkpoint (v5.2/6) when I was first starting out (it's still in the top 3), but I wish you'd update it with a v8. It has a special place in my heart. Even just a minor v6 rework/update, nothing too major or time consuming.
I know v7 got a lot of hate at the time because it was so radically different, & a lot of people weren't willing to learn using it (I thought it was cool), but I hope that hasn't put you off ever returning & updating this particular one....
Also, from one ADHD fam to another; I hope life is treating you kindly & things are going well ;) ^^ <3
Ever since I started to craft pictures on my own rig, this is still my favorite checkpoint. Thank you for this magnificent work. Creating boys and men with this model just hits on a completely different level.
Just beautiful. High quality output is always an invitation for further experiments.
i love this model and miss it.
Best model out here!
I don't get it. But I want to. Could you please tell me why you think it is? <3
kinda crazy that this is what I get after saying:
(african_west, golden brown skin, brown skin, dark skin, black woman)
neg: (european, asian, anime, pink nipples, white skin,:1.5)
and adding 3 skin tone loras
Why can`t I use this checkpoint on the platform anymore?
Details
Files
Available On (3 platforms)
Same model published on other platforms. May have additional downloads or version variants.