Yes, absolutely. Drove me mad for a couple of hours, but forcing it to use 3.10.13 (in my case) fina
Yes, absolutely. Drove me mad for a couple of hours, but forcing it to use 3.10.13 (in my case) finally did the trick. Love your SUPIR work Furkan!
Fthank you so much
Zi was told that supir would work if i had enough system ram, so i upgraded to 64 and comfyui is still saying out of memory, do i need to run comfy in low vram mode?
Z
Fwho said that?
Fyes it will work but if it uses RAM instead of VRAM it will be like 20x slower
Fso you have to wait
Fwhat is your VRAM?
Z12 gb
Zand 64 gb of system ram
Fok if you dont upscale too much
F12 GB will work
Zbut still said oom
Fah
Zdo i need specific arguments when i run comfyui to tell it to use system ram
Fi dont know comfyui implementation
Fours works
MI am wondering if this technique could be used to upscale images by having high res reference photos https://realfill.github.io/
Like having an old image in front of a church but the church still looks the same, so you give it a today taken photo as reference to enhance the photo. Or you want to colorize an photo of the original Titanic, so you give it photos of the movie Titanic where they restored it by hand
Like having an old image in front of a church but the church still looks the same, so you give it a today taken photo as reference to enhance the photo. Or you want to colorize an photo of the original Titanic, so you give it photos of the movie Titanic where they restored it by hand
RealFill: Reference-Driven Generation for Authentic Image Completion
Fi think this is generative fill
Fcontrolnet already have this
MReference based? I don't think so
Fby reference arent they meaning original input image expanded?
MHave you read the paper? It's about training the AI on reference pictures from a simular scene with at least one image as training image to expand or inpaint an target image for a accurate representation.
MStable Diffusion plugins just recreate the images or inpaint randomly by it's text input. But RealFill uses actual reference images that are accurate to the target image, hence why it's called RealFill. Like in my mentioned examples above, this could be very useful for restoration.
MI thought maybe this could be transformed to be used in upscaling by giving a high res reference image from the same scene/subject
Fi didnt read yet but i have a feeling that image is used as prompt
Fthus it wont be loyal to input :d
Mwdym? That's the best case scenario you could have when you are allowed to give the AI the information to make an accurate restoration of an image or video or what ever instead of letting the AI think what it sees.
MThere was a video from TwoMinutePapers about it
F2 minute papers never uses himself just show the demo which are usually very misleading 
Fbut i can make an auto installer for this
MYea that's why Two minutes 
MI thought i mentioned it here because maybe it could be used in SUPIR? Or something else?
Fmaybe need testing
Zis there a better version of xformers for comfyui than 0.0.24?
Fxformers 0.0.24 is latest
NHi , is there a working model to detect Deepfake images/vids?
Fnot that i know
Z@Furkan Gözükara SECourses Has there been any kind of SUPIR implementation into automatic1111 or webui-forge yet?
Fi dont think so
ZZI see an sd21 and sd 15 model so far
Fwhat was media pipe doing?
Fi think my auto download has all the available SDXL controlnet models
ZMapped facial. Expressions basically to do non ordinary smiles
Fi never used sadly
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
DI see facefusion has a new update, however im getting an error - RuntimeError: D:\a_work\1\s\onnxruntime\python\onnxruntime_pybind_state.cc:743 onnxruntime::python::CreateExecutionProviderInstance CUDA_PATH is set but CUDA wasn't able to be loaded. Please install the correct version of CUDA and cuDNN as mentioned in the GPU requirements page (https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements), make sure they're in the PATH, and that your GPU is supported.
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=16)
Fi am not using at the moment
DWill updating the cuda and cuDNN to the latest versions break anything?
Fprobably since it uses onnx
CI would like to ask for some help. Have you known a comfyui AI that can create json files based on a prompt (or/and pic or/and video) that can be dropped into comfyUI? If not, how would it be worthwhile to start making one?
