Inpainting in Comfyui | Inpainting models

Inpainting in Comfyui | Inpainting models | AI art generator

Рет қаралды 3,456

Күн бұрын

🙂‍ In this video, we briefly introduce inpainting in ComfyUI.
🤔 When inpainting images, you must use inpainting models. Standard models might give good results, but it is only a coincidence.
🫡 By the way, inpainting is very difficult to control.
👉 Get my personal ComfyUI inpainting workflow template at this link,
bit.ly/3ve0Bq8
🤓 The video tutorial covering how to get the starting image for the inpainting step,
• ComfyUI img2img workfl...
Chapters
00:00 Introduction
01:05 Model cards - Very important
02:32 Civitai - the marketplace
03:40 Let's do some inpainting
04:48 How to mask the image
05:31 Remember to use inpainting models with inpainting
06:52 A first result
07:39 The influence of denoise
08:33 Let's auto-queue until we get a nice result
09:41 Bingo, a lovely composition
10:30 Let's add sunglasses
14:31 Let's do some upscaling
18:54 A final inpainting example
21:23 Hello Pikachu
22:43 From this point, the inpainting was not working
26:07 Final remarks
Inpainting models:
huggingface.co/stabilityai/st...
huggingface.co/runwayml/stabl...
huggingface.co/diffusers/stab...
huggingface.co/Lykon
civitai.com/models/4384?model...
civitai.com/models/4201?model...
Beginner’s guide to inpainting (step-by-step examples)
stable-diffusion-art.com/inpa...
ComfyUI
github.com/comfyanonymous/Com...
ComfyUI Community Manual
blenderneko.github.io/ComfyUI...
ComfyUI Workflows
comfyanonymous.github.io/Comf...
github.com/cubiq/ComfyUI_Work...
github.com/wyrde/wyrde-comfyu...
github.com/atlasunified/Templ...
comfyworkflows.com/
Custom Nodes, Extensions, and Tools for ComfyUI
github.com/WASasquatch/comfyu...
Other links:
civitai.com/
huggingface.co/
creator.nightcafe.studio/
lexica.art/
****************************************************************************
Wolʇ Dynamics, your reliable partner for CAE solutions,
CAD and solid modeling → Meshing → Simulations → Automation and optimization → Post-processing → Data analytics and ML → Reporting
www.wolfdynamics.com/
Why Wolʇ Dynamics? Simply look at our banner - notice how the F appears backwards.
Wolʇ Dynamics ↔ Flow Dynamics 😯+💣=🤯
****************************************************************************
👉 Subscribe and hit the bell to see new videos:
/ @wolfdynamics5703
🖖 Join our channel to help us create more content. By joining our channel, you can also get access to perks:
/ @wolfdynamics5703
Follow us:
Twitter → WolfDynamics
LinkedIn → linkedin.com/company/wolf-dynamics
#ai #aiart #aiartgenerator #SDXL

Пікірлер: 11

@blacksage81 7 ай бұрын

There is a simple recipe online to turn any model into an inpainting model, the only issue is that one needs enough VRAM to use Auto1111's checkpoint Merger tool. The recipe goes like this, Model A + Model B + Model C Multiplier of 1, Where Model A is the SD base inpainting model, be it 1.5, 2, or XL. Model B is, the model of your choice(it needs to be of the same version 1.5, 2, XL). Model C is the SD Base model of the the same version as the previous two Models

@WhatsThisStickyStuff 4 ай бұрын

For some reason when i apply the lama model for object removal nothing happens even though I have the same setup as you

@wolfdynamics5703 4 ай бұрын

do you have the same checkpoints? remember that sometimes you need to give a prompt to erase the object, something like same background

@marksrye 5 ай бұрын

unless im missing something the workflow link seems to go off to a png and not the json file for the workflow?

@wolfdynamics5703 5 ай бұрын

you can drop the png on the canvas and it will open the workflow

@godofdream9112 6 ай бұрын

drawin to img in cumfyui... i dont have a1111.. also a1111 is slower than cumfyui in my pc.

@wolfdynamics5703 6 ай бұрын

Yes, A1111 tends to be be slower than comfyui, but there are ways to speed it up. Hardware is also important, with less than 8GB VRAM, A1111 will be slow. Another comment, there is no official support for LCM Lora and SDXL turbo.

@user-WernerF 6 ай бұрын

With all SDXL Inpainting models I get this error: Error occurred when executing CheckpointLoaderSimple: ERROR:root:!!! Exception during processing !!! ERROR:root:Traceback (most recent call last): File "F:\KI\ComfyUI_windows_portable\ComfyUI\execution.py", line 155, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) aso. 'model.diffusion_model.input_blocks.0.0.weight' With SD1.5 Inpainting models it works fine. Why?

@wolfdynamics5703 6 ай бұрын

Hi, you need to use an inpainting model for SDXL. You can try this one huggingface.co/diffusers/stable-diffusion-xl-1.0-inpainting-0.1 It works with a1111 but I havent test it in comfyui

@user-WernerF 6 ай бұрын

This one and all other SDXL-Models work only as second model in a workflow for the inpaint part but not as a single loaded model@@wolfdynamics5703