If you enjoy my contribution to this community, feel free to buy me a coffee: the more caffeine I drink the more models I can create š
Comparison gallery here: Fv6-Fv5 and NSFW illustration comparison between Gv1-Cv6-Fv2.

Current SOTA model in my experiment:
SDXL Base model: Gv4 is the most balanced model allowing both realistic and styled NSFW and SFW images. Better aesthetic than Fv6 but less photorealistic.
SDXL photorealist (SFW and NSFW) model: Fv6 is the way to go for hyper-realism including realistic NSFW images but it mostly lacks the styling capabilities of Gv4.
SDXL Illustration : Gv4 (SFW and NSFW). Cv6 is however still worth a try if you are not into NSFW images.
Flux model: Flux1-A1
Image Generation Settings for SDXL models
DPM++ 2/3M SDE / Karras or Exponential are always a good bet with 25+ steps and CFG around 5-7. But DPM++ SDE / Karras with less steps (e.g. 12) and higher CFG (8-11) is worth a try as is Euler Ancestral / Normal for a less detailed result which may also be interesting with a model like Fv6 to get something less "noisy".
The default CLIP Skip of 2 is also a good bet, but using 1 or 3-4 is also worth trying: 1 push more towards prompt adherence and 3-4 give sometimes a better result than the default focusing more on the "concepts".
Image Generation Settings for Flux models
My preferred settings are DPM++ 2M / beta or sgm_uniform or DDEIS / normal for the sampler / scheduler, beta giving a bolder stronger image. For a more subtle image, Euler / simple or beta seems a good bet.
CFG seems to have a huge impact on the final image and be very sensitive even to small variations.
For photos, CFG should remain low (1.5-2.5) to avoid plastic skin.
For fine art and illustration it is more complicated because it depends on the medium. For "rough" styles (painting, watercolours etc.), CFG should stay quite low in the 1.5-2.5 range but for anime or comic style, CFG needs often to be pushed further to achieve the desired style (3-6 or more).
If the image is messy/malformed or blurred, it is often because the CFG/steps are inappropriate for this image, but it is not always easy to know whether CFG/steps must be increased or decreased (at least to me š).
There is for sure a lot to learn concerning Flux behaviour which is quite different than SDXL and we will need to adapt.
Workflow
All my images are uploaded with embedded ComfyUI workflow which is alas incompatible with CivitAI processing and most often prompt cannot be retrieved. You can however download the original PNG image with the workflow included by clicking on the "DOWNLOAD" icon in the image viewer.
Why I try to mostly publish images straight from my model with maybe a bit of a Lora (mine or some detail enhancer), I also sometimes now use Controlnet to get better more detailed compositions more easily: in this case the source image is obviously not in the workflow, but I guess you can still use the image I publish as a source if you want to make a variation :)
Past
Starting from E serie, models are evolving sometimes by merging with other models (thanks to other contributors!), but mostly via training on my own dataset: a modest dataset (~2000 images currently), but I try to somehow compensate with quality and originality.
Starting with Fv1, I have included many synthetic images I created using previous versions: playing hard with prompt and retouching when necessary the result in Photoshop in order to have a dataset which contains many original images.
The core idea behind this model was to create a versatile tool by merging some of the best existing models which fits my personal taste (photography and fantasy art to make it simple). My primary goals were:
Photorealism: The ability to produce stunningly realistic images of both people and objects/nature.
Flexibility: The ability to create highly stylized images, allowing for artistic expression through various styles and combinations of artists. I am from a older generation and comes from Europe so "style" does not mean for me "Japanese kawaii aime with boobs" or "DC Comics cartoon with lots of superheroes and voluptuous blonde babes" but more from a univers of Frank Frazetta, Milo Manara, Boris Vallejo, H.R.Giger, Wojtek Siudmak and such fantasy art masters: there are boobs involved for sure, but the style is somewhat different :P
As I doesn't like being limited in my exploration of the human body, the idea is also to have a fairly capable NSFW model. However, dur to the nature of available training images in the data sets, NSFW often comes with a strong bias toward either porn photos or porn Japanese anime and impact the flexibility (typically as soon as you use the word "sexy" in your prompt you need to weight in the style). This point is thus NOT the priority for the base model but is pushed forward in the NSFW model.
Description
FAQ
Comments (14)
Hello, I really like your model, what advice do you have for inpainting with your model ?
I really like your model, but it performs poorly with male nudity. The penis often appears deformed.
Thanks! I guess for some reason training datasets contains more female genitals examples than penis :P Maybe some Lora would help on that point?
this model's on another level. First one I come across that can do water refraction.
Thanks!
Water refraction was not on my to-do list, but glad it does it well :)
Araminta Gv4 is in my all time top3 list of checkpoints. <3
Yes, since mentioned the checkpoint in another topic, reminded me to come here also to show appreciation for this amazing checkpoint.
Love the aesthetics and on top of that it CAN be very well-round "all purpose checkpoint" if enjoys the unique style it has.
I do like your other checkpoints also, Fv5 is amazing too. Gv4 just has that special something, extra artistic emo powers can fully go of the charts when using it. ^^
Your experiments are much appreciated, so my humble thanks to the creator for your work. <3
Hey, thank you for your kind words :)
I agree that while Fv5 is better at realism, Gv4 is a far better all-in-one model.
@aramintastudioĀ Yeah. :) I think Gv4 is at a point, where it's sort of at a perfect spot as is, since it's a good combination of aesthetics and realism. :) With Fv5 can probably do still a bit experimenting, even though it's like 95% perfection. :D
Fun fact, i did some funny idea of photorealistic fantasy creatures/aliens/mythical monsters in everyday situations, like a zombie washing the dishes in a Scandinavian apartment etc just for example of one scenario lol, so i used many different Loras, like with zombies, monsters, aliens, like Na'Vi from Avatar movie etc.. so i tested about 10 different checkpoints, since used like 2-3 Loras at same time for maximum realistic results, and in your Fv5 actually handled that with best looking results, all were skin effect Loras with eye, hair and other details that can think for fantasy creatures. So Fv5 was actually best in that.
Fv5 in general has STUNNINGLY good skin and eye details (top3 tier thing you have baked), if makes 4K images and uses the SmartSharpen thingy node in right settings, the results are insanely good even without a single Lora for normal human characters, but with Loras it was amazing how well it meshed out for fantasy creatures and results looked like actual fantasy creature eyes and skins, like super organic with tactile detailing. Actually impressingly good checkpoint. The eyes, you know especially the irises stroma details, could zoom in and they look so real, like miniature galaxies with unique patterns like in real human eyes irl (or in this prompt simulation we are living in :P).
So yeah, once learns how to prompt both Gv4 and Fv5 to own desired results, they are both amazing. Gv4 for slightly more artistic/aesthetics direction and Fv5 for pure realism. "Hidden gems" type of things, if i am honest, these should literally be in the top3 most downloaded checkpoints here in Civitai, up there with the big names. ^^ So keep it up!
Was trying to think if there is anything to improve on for future checkpoints, only could think of one, sometimes when makes petite emo/goth characters, their lower body, especially from side or behind angle in cinematic shots can look a bit on the voluptuous side, when obviously tries to go full petite & skinny look for artistic reasons, so then has to fight that happening with prompt gimmicks and changing angles, might be my personal skill issue tho.. :D So it can look like 2 different persons body mixed in one, but just from certain angles. Nothing wrong with big booties tho. :P Just that gotta match petite upper body with petite lower body. Think that happens with many checkpoints in general tho, some perspectives makes them have distorted details, that are hard to correct with prompting. It affects clothing also, i like making clothed SFW images, like gothic/emo clothing etc, it's like trying on cute clothes, without having me to physically do it myself lol. So i guess that is only thing i can think of, and even that is not your checkpoints original sin, think it's something in most of them or in SDXL in general, the perspective distortion in some angles.
Enough rant, sorry. :D But if you can fix that small issue and improve upon Fv5 for that added realism of body shape staying consistent in all angles, your checkpoint would be legit best one there is. Gv4 is kind of at the top of it's game already probably, cannot improve perfection.. i guess? :P
@EclipsiaĀ Wow... you definitely tested my models better than I did :)
Thank for your feedback... not sure if/when I will try to make a Gv5/Fv6 version (seems we are reaching the limit of SDXL fine-tuning), but I will keep your petite-bodied points in mind :)
@aramintastudioĀ Hih thanks! ^^
And yeah, i'd even take a really well done Lora for petite body types if some creator makes one, but would be cool if there would be a checkpoint or two that would have such baked in naturally. I mean.. you know how many checkpoints and Loras go the exact other way, of bringing SUPER voluptuous and huge breasts and big wide hips and all that.. nothing wrong with that, i like making images and subjects of all shapes and sizes, but i feel easier to often make "giga babes" with big butts and massive breasts, than petite-framed ones. I like when easy to use full variety of choices. Human body is beautiful in all shapes, sizes and forms. ^^
As a special note, i REALLY like your checkpoints anatomy and aesthetics flexibility. Also, that it can give very surprising (in a good way) results if leaves some things open in the prompt. So gets many different kinds of people and body types with same prompt, what makes things interesting for us who tried many checkpoints, some other checkpoints have way more repeating feeling in results. I like when gets some total surprise details in the same scenario with each seed change. And your checkpoints have way more natural breasts etc than most (you know, many have quite "Playboy silicone boobs" types, again, not saying it's wrong to those who enjoy them hih, but nice to have all options open for any type of looks, and how they react with clothes, like tops/camisoles etc. So more artistic feel and interesting images are often possible with your checkpoints.
So your checkpoint is for the true culinarists of aesthetics and appreciators of human body. ^^
And yeah, i understand and seen, SDXL been pushed quite close to it's limits, some checkpoints every now and then add nice finetuning still to certain areas, but yeah.. often now Loras make bigger difference to pin-pointed image ideas when combines to your favorite current best checkpoints. Hih again i ended up ranting, was just supposed to answer quick lol. Anyways, have a great summer weekend and thanks again for your amazing checkpoints. š
@Eclipsia I completely understand your pointā1000%, in factāregarding the biased body features often found in many models (oversized breasts and buttocks, with barely any body hair in sight). Itās an issue Iāve been trying to address since my very first A model.
My models are the result of a careful mix of merging and fine-tuning, using a custom dataset composed exclusively of people who are ādifferentāāmore realistic, more diverse, and (in my humble opinion) more beautiful, with higher aesthetic value. Of course, beauty is always a matter of taste.
Iām (or was?) a photographer, and Iāve always tried to push aesthetic quality as far as I couldāhopefully that comes through in my models.
Also, I have a preference for petite body types, so I hope my models reflect a bit more balance on that front too :)
I realize that sdxl is the closest thing to creating realistic images, but I've tried a bit of KREA, and being a loyal follower of your work, do you plan to create a checkpoint based on KREA?
I don't have much time to invest in other base models right now and there are too many new models coming out right now (Qwen Image just today) to choose one.
I guess I am currently more in a wait and see period :)
aramintastudioĀ and chroma version 1hd is out already as well soo.š