I rather FML now, it's fantastic. This version is producing faces with biased formats, like very oval or circular.

hi there. those generated pictures samples are describe of 'epic realism' and really 'i can't believe it's not photography' also it's 'photomatix'ally beautiful and 'realistic vision' of 'absolute reality' in shape of how a just 'lazy mix' should be and shouldn't in terms of 'analog madness' of a 'perfect world' 'photon' can travels through 'azovya' to produce a 'picx_real' generated pictures using the AI model stable diffusion version 1.5 in 2024.

With such a diverse model how do you accommodate different categories (cars, people, etc)into your training or dataset preparation?

1. Do you separate categories into different folders or do you put all the images into one dataset? And how important is captioning?

2. What's the secret to training hands well?

How the hell is this model so clean and coherent? Just ridiculous O_o

@epinikion

Thank you for releasing so many realistic models.

I would like to quantitatively understand the performance of each model. I created a table based on the information written in [About this version].

Do the numbers in this table match your feelings? Please let me know if there are any corrections.

For photorealistic, [AmateurRealLife] has the highest score of 10.

For versatile, [Ultimate Fidelity] has the highest score of 10.

The table is based on these maximum 10 points.

photorealistic versatile

Ultimate Fidelity 8 10

Last Unicorn 8 7

AmateurRealLife 10 4

Z - Universal 9 8

Y - Photoreal 8 7

X - Photoreal 7 7

X Plus 7 7

X Plus Plus 7 7

V4 - One 4 All 6 6

V3.1 - GetWhatYouPrompt 5 5

V3 - GetWhatYouPrompt 5 5

V2 - Dodge and Burn 5 5

V1 3 4

Anyone have advice on creating a consistent character? Don't need a full explanation just where to start (things to download, specific settings, etc)

needs some adjustments to improve the quality of the hands, maybe using nail polish images can help improve it. Use some very detailed hand images

By far the most consistent SD1.5 model I've used. Highest chance of accuracy on feet and hands, good body symmetry and probably imparts the last amount of arbitrary or overfitted "style" to faces.

Out of curiosity, any plans to create a SDXL- or Pony-derived model?

This is easily THE best realism model on this entire site. The only one you'll ever need. I've experimented with dozens of the top rated "realism" models. And they all fall short of the results produced by the Ultimate High Fidelity model. This model is truly a work of genius.

If you have a good GPU, I'd recommend going with 100+ steps. And using the 4x-Ultrasharp upscaler as your HiRes fix. It also seems to work extremely well with the "more details" lora at around "0.7". You'll be able to zoom in and see veins, stretch marks, and pores. The "face_yolov8n" ADetailer is also a must.

epiCPhotoGasm has more restrictive license than epiCRealism, having "no sharing merges", etc. Yet there are models on Civitai that merged epiCPhotoGasm. Is there a list of creators or specific models that got permission from epinikion to use epiCPhotoGasm as one of their merged models?

Been using Last Unicorn since Christmas 2023 for photographic images I've still got to find something which outdoes it. Also does very well at everything else I've tried it on. I have had some incredible images out of it.

Hands seem to be an issue with almost all safetensors and this does the usual extra fingers, less fingers and distortions. I doubt if that is a problem with the safetensor but SD itself.

If I want to create a LORA that works best with this checkpoint, what should be the base model I use to train my LORA with? Thanks in advance!

Need to improve on the tits. Nothing but G cups in here. o.O

I've a Epicphotogasm Model ID .. But cant find which model it is ? Could any one help please?

Model ID is fe06753eee

I was so confused why the results are so bad and people face LORAs look so weird and different from the real person. Finally found out that it's an issue unique to this Fidelity version. I reverted back to Z Universal and it works perfectly.

Great model, excellent job at picturing reality. But, as like most other models and LoRAs, faces come out weird.

I want to start by appreciating all the work put into this model. By far, the "Last Unicorn" version is the best of this model. Unfortunately, there is a big problem with the generated colors. Very often, the colors on the clothes are applied to the environment. Also the material the clothes are made of is somehow added to the environment. So far I have not been able to prevent this problem from occurring. Perhaps in future versions, if there are any, a solution can be found.

This model is insanely great. I mean, THAT great! I was successfully integrated my LoRA into this model, after so many trial and error. This model was fine-tuning into the realism of the photograph. That is the challenge. If your LoRA is "not good enough" and has poor quality, it won't make it. I was able to create a realistic person image with this checkpoint. Initially, I used to have the Beautiful Realistic Asian V7, ChilloutMix, and MagicMixRealistic as my checkpoint. But this one has topped it all. Ultimate Fidelity is my favorite. The solid checkpoint!

Awesome model! I used the Last Unicorn Version a few weeks ago. Unfortunately I cannot use It anymore tongenerate Images in civitai. It says "Unavailable resources". But I see that Images are generated with the Last Unicorn Version a few days ago. How can I use this model?

I have found a site that copied many models with no generation services rule set. here is the page with your model. https://88stacks.com/models/KVe3qCZEL7pBo--epiCPhotoGasm o

Can you please post the ultimate fidelity inpainting model? Some don’t have the ability to merge models to create it. Thank you

I hope you can make EpicPhotogasm in Flux version

This model is certainly wonderful. It reveals itself as effectively as possible in close-up portraits. Detailed skin texture, film grain, artistic photographic lighting and colors..., all as you like. On medium and wide shots, it also shows itself quite well, the same good photorealism and good background detail, although sometimes glitches of anatomy may occur.

Perhaps this model sometimes lacks a little responsiveness and flexibility, but in photorealism it is really at the top of Olympus. Great job!

For an SD 1.5? Unreal. Sometimes i have to compare it to Flux. Details, sure but... more flexible. Its tough... and this is first gen (well 1.5). Flux is what? 4th 5th gen and much heavier?

why no fp32 version uploaded?

Why this model is so good ? getting better results than Flux while using 1/8 of the PC resources and time.

Hello. I have best results always with X plus plus. But when using Unicorn, outputs are not very clean. Why? All comments say, that ouptuts are aweswome? What could be the reason?

THE photorealism. Best of the best, outclasses any other, it's just ultimately perfect in that regard.

Please make an XL Lighting, Turbo or Hyper version.

This is the best model for photography.

very nice!!!!!!

There is still no XL/Pony model that can rival this for realism. This generates beautiful, photorealistic bodies. No XL/Pony model has matching quality. Unfortunately, being based on SD 1.5 means the prompt adherence is notably worse.

how do i get workflow for inpainting?

We used epicphotogasm_lastUnicorn Version v1.6.0 in a university course very happily. This year we updated to v1.9. The new version started to produce nude people even if not asked so, e.g. for the simple prompt 'several women'. That was quite startling and inappropriate. I think you need to check that, because that's not acceptable for normal use.

For me, it's the most versatile 1.5 model. It can create very realistic scenes with wonderful, dramatic lighting. There's also a model named Analogmadness that is comparable but much stricter in capabilities. Also, notice that it's very strong at creating hypernetworks it's my go-to

Hello, I'm sorry for bother you all, I'm new in this AI world and I would like to know how I can use this model, thanks!!

Like the realism, but unfortunately, most of models from this creator cannot create black people without over-sexualizing them. Any prompt or even just "black woman" returns sexualized images but doesn't happen with others groups. Very unfortunate.