Recommended Settings:
DPM++2M - SGM Uniform
2.7 Distilled CFG
LoRA Weight 0.5 - 1
20 - 30 Steps
Feel free to test other settings and comment on them.
---------------------------------------------------------------
We have several models that make images more realistic, so why am I launching this one?
Unlike the other models, my goal is to bring more realistic textures and lighting, rather than a boring conventional photograph.
The effect sought is as if it were taken with a professional camera, but without that painted look that the base Flux brings, even when using terms in the prompt to achieve realism.I uploaded a collection of images comparing the same prompt and seed, with LoRA (right) and without LoRA (left).
Notice the texture and light details that the model brings to the image.
And this isn't limited to real photography and people; it can be applied to creatures, landscapes, animals, abstract art, and whatever your creativity allows. It works very well when combined with other LoRAs.
I am very curious to see the possibilities of this model, so please share as many images as you can here and give me feedback so I can further improve the model.
Description
FAQ
Comments (25)
This LoRa has the approval of the Devil himself.
Thank you, Devil
@evertonlisboad367 Oh, give it a rest. Being nice to me won’t save your soul, boy. When your time’s up, it’s up no charm or sweet talk will change that. But hey, nice LoRa, I’ll give you that.
I love the effect but it does tend to skew men, at least, towards skinnier and lankier than the model produces without the lora.
Oh, that’s because we have no business cheering for body positivity when it comes to this. Being obese isn’t some cute little trend to celebrate it’s a ticking time bomb, a real dangerous health issue. So, let’s not pretend otherwise, shall we?
@ELDiablo I'm referring to turning a "buff, muscular man" into a twink, but you read into the original post whatever you want, mate.
@ELDiablo "this is why a generalized lighting & detail lora skewing towards a specific result is good, actually" lmao. come off it
Works as promised, maybe too well, have to try dialing it back a bit :D
This is great. Gives a good amount of realism at even lower weights of .25 or lower.
Hey thanks for this it works reallllly well!
Hi friend, just popping in from Reddit to mention that you should ignore them all, the LORA is fantastic. They don't know what they are talking about, like 99.9% of the time on that festering rotten website. Their pleb downvotes can't reach us here. ✌️
I really liked it! It gives a professional look and fixes some of my other LoRas that accentuated the saturated colors the model is prone to do sometimes. Thanks!
After I finished licking my bedroom window I finally got flux loras to work, that was very hard btw, because not enough people are sharing information about how to connect the noodles, correctly. This lora actually makes images look more realistic. My image I uploaded here is the Crystals in a forest, this lora is impressive.
I'm completely new to AI image generating. Using ComfyUI since the nodes work flow looked tempting (an old Blender user). Yet, I can't find many node tutorials for Flux and the other models seem to be using different nodes. Do you have recommendations? I already struggle to get the LoRA weight fixed.. xD
I saw your post on Reddit, your lora improves the texture quality considerably. Just wanted to say thanks :)
Forget the Devil's approval, you've just been blessed by God Almighty. Good Lora in the jungle of useless Loras.
How to get even better results with this LoRA: https://civitai.com/articles/6992
It doesn't seem to work with the full fp16 version of Flux Dev for me in comfyUI , it throws an error at the KSampler, which is a shame as I think it is really good when using it with fp8 flux.
Ahh, I figured I am likely running out of Vram when loading the lora as well as the fp16 model on my 3090.
All the Loras released since and this is still by far the best!
Could you please share the full config file used for training? this is really good work nice.
this looks so epic, not gonna lie!
Im trying to use this model but comfy gives me this error, ERROR: Could not detect model type of: C:\pinokio\api\comfy.git\app\models\checkpoints\Phlux.safetensors . Someone could help?
that's a lora, should be put in models/loras, not into the checkpoint folder, it cannot be used aass a full checkpoint, check some tutorial online on how to use loras if you need to
Brilliant Lora. Kudos. In B&W realistic photography is unparallel. BUT gives me B&W only half of the time. The other half its desaturated colour. How to make it consistent in B&W?
Details
Files
Phlux.safetensors
Mirrors
Phlux.safetensors
Phlux.safetensors
Phlux.safetensors
Phlux.safetensors
Phlux.safetensors
Phlux.safetensors
Phlux.safetensors
Phlux.safetensors
Phlux.safetensors
photorealism.safetensors
Phlux (1).safetensors
Phlux.safetensors
48-Phlux.safetensors
【风格增强】【FLUX】Phlux-Photorealism with style.safetensors
1122Photorealism01.safetensors
phlux.safetensors
Phlux.safetensors
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.