#ComfyUI has a plugin that can pip install new python packages, update them and keeps its registry of addons in Github. Would be nice to see something like this in #plone

I made my first custom nodes for #ComfyUI. They're for computing optical flow between two images (generally adjacent frames in a video), visualizing the flow, and applying it to another image (generally the previous output image in a #vid2vid workflow).

https://github.com/seanlynch/comfyui-optical-flow

#StableDiffusion

Simplified my SDXL Workflow organizing it in a way that controls and results are closer to each other. I got tired of dragging the screen a lot to get to the results.

#comfyui #stablediffusion

ET enjoying a big mac
#aiart #stablediffusion #ComfyUI

If you want to regenerate this image you need #ComfyUI itself (https://github.com/comfyanonymous/ComfyUI), these controlnet-preprocessor nodes (https://github.com/Fannovel16/comfyui_controlnet_aux), SD XL 1.0's base (https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/tree/main you need sd_xl_base_1.0.safetensors) and refiner (https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/tree/main you need sd_xl_refiner_1.0.safetensors) model

(1/x)

Sigh. Unfortunately it looks like Mastodon also no longer saves the metadata of images. At least it doesn't save #ComfyUI-workflows. Ok, this image is actually one of two images I submitted for a contest on CivitAI (https://civitai.com/posts/543705) and from there the workflow can actually still be copied to be inspected in ComfyUI.

To all anti-AI-artists who claim "AI-art is just typing in some prompts.": here is an AI-generated image and the custom #ComfyUI-workflow I used to generate it.

I know that using AI to generate images is a touchy subject, but I'm having a lot of fun using ComfyUI to generate artwork with Stable Diffusion XL for the next Synth You've Been Gone YouTube video.

The node UI is pretty intuitive and powerful, and the images it comes up with are great.

#sdxl #comfyui #ai #synthyouvebeengone

#SD-XL runs on my crappy 4GB card using the lovely #ComfyUI (node-based...like #Blender!)

Man leaning looking over cliff.
#aiart #stablediffusion #SDXL #ComfyUI

#AIArt Zoom "Flood"
#ComfyUI #StableDiffusionXL #SDXL

So sieht das ganze Endprodukt dann aus :)
#ComfyUI #AIArt #SDXL

Poll has ended and what I take from this is:
1- People want to learn how to use ComfyUI
2- There is interest in training LoRAs
3- Animation using Deforum is not as popular as it used to be
4- People who voted "Something else" couldn't say what else they wanted........

#ai #aiart #StableDiffusion #ComfyUI #lora #a1111 #sdxl

Animated zoom - MK2
Created with #SDXL #ComfyUI #AIArt

Nice to see #ComfyUI is now using MPS on macOS M1/M2. The initial release had to use the CPU and took over 3 seconds per iteration. Now it's getting 3 iterations per second, for a 10x speedup! #stablediffusion #aiart #macos

Animated zoom.
Created with #SDXL #ComfyUI #AIArt

ComfyUI is… interesting. I feel like the concept of nodes has a lot more potential power for flexibility than how Automatic1111 presents, but wow… if Automatic1111 wasn’t complicated to figure out how to use effectively, ComfyUI sure is.

Honestly it should come with several pre-built workflows for img2img, inpainting/outpainting, upscaling, SDXL, etc.

Still can’t get a plug-in to work well with Krita though… that would be great.

#AIArt #stableDiffusion #comfyUI #automatic1111

#StableDiffusion #AIArtCommunity #aiartwork #sdxl #ComfyUI #birdart #MastoArt

Going to make some owls here.

#StableDiffusion #AIArtCommunity
#aiartwork

I've got #SDXL up and running in the official #ComfyUI workflow from Stability.

God partial, big bang
#airt #stablediffusion #SDXL #ComfyUI