i am waiting next gen gpus to move to 128 gb faster RAM
i am waiting next gen gpus to move to 128 gb faster RAM
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
Dwhat is the difference between comfui installer v3 and v4?
Dapart from the obvious 
AWhats hugging face version mean?
Fi mean diffusers : https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/README_flux.md
GitHub
Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch and FLAX. - huggingface/diffusers
Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch and FLAX. - huggingface/diffusers![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=16)
Fi dont remember check info 
Dlol
Ais there a downside to simple tuner?
Fharder to use. i dont know if working accurate never tested.
VShouldn't that be expected?
VWhen even 24gb dedicated ram becomes less
.
.VDoc we need to do a GPU heist at the Nvidia headquarters .
.
Fwe need
NThank you so much for this, you've optimised one of my processes around image processing so much with this!!!!
NI don't think 5090 will have that much VRAM, 48GB at max, I dont even think nvidia will go that route. Truly cloud computing for this stuff is the future-- and I am saying this as a 3090 and 4090 owner 
F48 gb would be amazing
Frumos are like 28 32 which are totally lame shameless
Fyou are welcome. this is the best atm
NThese examples are extremely good (at work so can't test). Was this using the general model or the human specialised model? https://huggingface.co/datasets/schirrmacher/humans
BI'm making a lot of grids for Very Special characters, so I built a quick web app to speed up evaluating them! I hope you find this free tool useful. https://everlyheights.tv/tools/stable-diffusion-tools/everly-heights-xyz-grid-evaluator/
Everly Heights | Stories set in the fictional town of Everly Heights, OhioEverly Heights
The Everly Heights XYZ Grid Evaluator is a powerful web-based tool designed to help you evaluate and rank outputs from your Stable Diffusion AI image models after you generate an XYZ Plot/Grid to test various settings, including LoRAs (Low-Rank Adaptation) and more technical settings like Steps and CFG. Whether you're an AI researcher, developer...
KNeeded in ComfyUI
KAll these stand-alone "apps" are cool and all, but it's not really useful jumping between all the different apps Upscaling is fine though sice that's usually a last stage
Upscaling is fine though sice that's usually a last stage
Fhi all, do we have any comfyui workflow to train flux on a character (like using dreambooth)
KNo. Not yet. Demands to much VRAM so noone has put any effort into that
KFlux trainings is done on rental-GPU's for now
Fokay i see, too bad!
FPS: you can run any comfyui workflow on rented GPUs on replicate: https://replicate.com/fofr/any-comfyui-workflow
Mhey doc, any toughts on how can i make the supir and vton app serverless after i deployed into runpod ?
Hinteresting.. can you run lora models as well?
FIs there a way to save your state from Massed Compute? It's kind of annoying that I have to reinstall all the custom nodes and models I wanna use everytime I boot an instance. Runpod's nice that I can keep a storage instance for cheap/hr
IHi guys, there's any tuto for loras on flux?
Sguys i tried to use topaz but a toaster runs better than my pc .. is there any colab or free alternative to topaz video enchancer???
Ihow long its the video?
S5 seconds
Ihmm then something its odd
Ais there a one click installer for ai toolkit?
Fi plan kohya tutorial soon for flux training
Fno idea you need some programming skills
Fsadly not
Fbest is upload into hugging face
Fwhich i do
Fhopefully soon
Fthere is this app but only on windows i presume
FGitHub
Video, Image and GIF upscale/enlarge(Super-Resolution) and Video frame interpolation. Achieved with Waifu2x, Real-ESRGAN, Real-CUGAN, RTX Video Super Resolution VSR, SRMD, RealSR, Anime4K, RIFE, I...
KAnyone else having problems running the NF4 model in ComfyUI?
KI'm fully updated
KRuns in Forge though
KTrying to wrap my head around the differences between the weight types, models and VAE's.
I find that using the 327kb .sft file or using the 163kb .safetensor file produces results that are the same
Using weight type default produces way better images, but it takes longer. VAE-Decode takes tie with default. With the other ones it just takes a second
I find that using the 327kb .sft file or using the 163kb .safetensor file produces results that are the same
Using weight type default produces way better images, but it takes longer. VAE-Decode takes tie with default. With the other ones it just takes a second
KI find the Flux models to be highly unstable though. Sometimes it just produces blurry photos for no reason
KOnly thing changed between these three images is the seed that is incremented with one:
WTF is going on... ?
WTF is going on... ?
