The idea behind this model merge is to create a perfect mix capable of generating artistic photography: artful portraits and beautiful landscapes.
Both HotArt and HotArt2 as well as HotArt3 are great in their own way, try both with the same seed/sampler = then use the one according to your taste ;)
This rich Checkpoint/LORA mix series is an experiment that worked out very well ;)
For HotArt2 better hands - use both of the suggested resources (LORAs) at low ~0.333 level.
See below for more examples of generation capabilities 馃憞
Description
idk yet
FAQ
Comments (16)
Would be really helpful if you would tell us what models have been merged into it.
As an example:For me its almost a no-go if anything anime,3d,2d,semireal-nonhuman"real"people is mixed into a REALISTIC-checkpoint. (That does not mean i do not like checkpoints with an unique style - artistic"reality"-style)
Well I give it a try anyway to get an impression of it.And cos I just cannot resist this beautiful vintage images down there in the showcase. :-)
the purpose of this model was to create as realistic-capable checkpoint as possible (so no anime or similar style models/loras were not even planned to be used as mix ingredients) - and results are incredibly photoreal
but I was not able to test its capabilities as an artistic model: impressionism or expressionism paintings (yet)
@OliviaRossi聽Thx for your answer. lets c how it works,I hope that I remember to post some example images this time,as you benefit from it as a creator.
@mscorpion7285聽good luck
Eh, in my experience a lot of those things can help improve realistic models. More facial variance, poses, colors, and so on. You can reduce how much they appear using weights or [ ] or negative prompting. FOr example, a lot of Pony models that are realistic actually turn out even more realistic when using some anime/cartoon/painting LoRA's with lower weights.
The biggest benefit in my opinion is that facial variance increase A LOT and I can prompt things like "Hispanic" or "Indian" and actually get someone that looks like those races.
We have to admit that most models, I'd say over 98% of them, are not much better than the default unmodded SDXL/SD1.5 base model because they rarely ever train more faces. It's extremely rare to run across models that can actually generate a variety of faces and is one of the things I hope SD3 will help to address by introducing natural guidance to facial variance when creating people.
The world is filled with color and vibrance and I'd rather have some of that in my generations than muted tones that don't represent what I see in reality. But, when I want something that is undersaturated and dull I should be able to prompt for that too.
FOr example, the black lady with yellow makeup is incredible and one of the best things I've ever seen rendered by an SDXL model or any model for that matter (DALL, Mojl, etc).
@LetTheBassDrop聽馃憦馃憦馃憦
@LetTheBassDrop聽Well, this aged like fine milk lol.
@colinw2292823聽Wine, good old wine馃嵎
Do you have a certain vae that works best?
all my latest models have VAE built-in to minimize reports by newbies that something is not working with them, but personally - I love that VAE with fp16 error fix, it is uploaded to Civitai and Huggingface
@OliviaRossi聽Is that fp16 error fix considered the standard that is baked in with most SDXL models? And do you recommend leaving the VAE selector set to automatic in most cases?
Just heads up, above the model description there is a sort of warning message stating that this checkpoint recommends a vae.
@BilboTaggins聽yes, I did it on purpose, because I'm unable to answer questions about VAE all the time quickly and on-time
@LetTheBassDrop聽my experiments with VAE: different VAE files (even the Tiled VAE in ComfyUI and Automatic1111) DO CREATE really different outputs, so TESTing is always a good thing! Use the VAE that works best with your current desired style and favorite sampler ;)
the fp16 fix standard SDXL VAE has always been OK and not a lot of improvement or difference was found with other VAEs, but recently i discovered Sharp Spectrum VAE for SDXL and it's different enough that I use it over the fp16-fix version in certain situations. It does dark and high contrast images better.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.