To encourage more people to create and share with our community, I have created a very simple starting workflow for everyone to use as well as adding a walkthrough below of how get started. As well as rules to follow when posting to Inkbunny.
This guide is for Windows Users
It is highly recommended to have an Nvidia GPU with 8GB VRAM and 16GB of system RAM. If you have a GTX1070+, RTX2060(12GB)+, RTX3050(8GB)+, or RTX4060+ card you're golden and will have no issues. Cards under this limit may work, but speed will be drastically cut and might lock up your computer. Most work will be done on your GPU so the more VRAM the better. For systems with GPU or RAM under requirement, an SSD could help a little bit, but in the end an upgrade to or above will help you run best. Otherwise reading a book or going for a walk might not be bad alternatives. You won't be able to game well while you're system is generating.
ComfyUI includes everything to run minus models/checkpoints & LoRAs, but is completely portable and can run completely offline once setup.
Once extracted you will go to the run_nvidia_gpu file and double click. This will download any required files before opening a window up in your browser.
At this point only need to download your models which you can find and download from either Civitai or Huggingface. I highly suggest either indigoFurryMix or FurryRock models to work with. When finished drop them into your "checkpoints" folder which can be found under ComfyUI > models. ^ ^
You should be able to hit the refresh button in ComfyUI to make your models load into the program.
Finally download either the included .json file or hit the download button on the image and right click and save as. You will now go back into ComfyUI and hit the load button and point it to the file you downloaded. This should instantly pop you into my beginners workflow which is labelled, grouped, color coded, and includes a guide for switching between basic generation and image 2 image generation as well as an explanation of what the settings do inside the KSampler.
I look forward to seeing new faces join in with us in making dreams come true!
Just remember that all uploaded images require your workflow to be posted to current rules as well as the comfyui and ai_generated tag or ai_assisted tag if you're adapting one of your drawn images or using it to create a background for a drawn image. This can be done by clicking the save button on the right once you got the image you like.
If I've missed anything or if you have any additional non-hardware related questions, I'm happy to assist. ^ ^
Before models install is about 5.5GB. Then I have 7 different models that are 54GB of space, but i pick the bigger versions so there can be saved space there as well. You could get by with a 512GB, but I'd probably recommend 1TB+ if you game as well.
Before models install is about 5.5GB. Then I have 7 different models that are 54GB of space, but i p
Yep! It works kinda like a LAN game server, so it runs from your PC and all the files are local. Your model files hold all the information to create images :3
Yep! It works kinda like a LAN game server, so it runs from your PC and all the files are local. You
It could be moved into an external SSD, but I'm not sure if its dependencies are installed within. I'd have to check. But the program doesn't have an installer, it's just a 7zip folder, which is an open source .rar for better compression.
It could be moved into an external SSD, but I'm not sure if its dependencies are installed within. I
I went and did some research and everything to run the program is included in the file download. Models however you will have to source from the websites I mentioned.
I went and did some research and everything to run the program is included in the file download. Mod
I'm lazy and using Fooocus. I tried sd.webui before that, but it tends to create garbage without extensive prompts. I guess ComfyUI is in the same gategory?
I'm lazy and using Fooocus. I tried sd.webui before that, but it tends to create garbage without ext
In general ComfyUI will work similar to Auto1111. I've never used fooocus but I see that it uses a GPT2 ai bot to help with prompts. There is an extension for auto that can add that functionality and might be a custom node for ComfyUI, but I haven't gotten that far. This walkthrough is more for absolute beginners, if fooocus works well for you, I'd stick with it. ^^
In general ComfyUI will work similar to Auto1111. I've never used fooocus but I see that it uses a G
Thanks. I might have to learn ComfyUI or Auto1111 at some point, as Fooocus has a tendency to create messed up images with furry models, even with all Styles (prompt enhancers) off. So I cannot actually create anything NSFW yet. Also, FurryRock and apparently indigoFurryMix too are SD 1.5 type models, which doesn't even work with Fooocus.
Thanks. I might have to learn ComfyUI or Auto1111 at some point, as Fooocus has a tendency to create
I see it's SDXL only, but can use a SD 1.5 model as a refiner, So you might be able to use a NSFW XL model then refine over with FurryRock or indigoFurryMix. XL wise so far I can see FenrisXL and Pony Diffusion V6 XL. It's a shame XL never caught on as much as SD 1.5, but on the bright side we got a ton of models this way and then Stable Cascade came from it, which I have high expectations from due to the higher sample resolution, lower generation time than XL or SD with similar size, and way easier text generation.
I see it's SDXL only, but can use a SD 1.5 model as a refiner, So you might be able to use a NSFW XL
I have to test more with the refiner switch, but usually it makes things worse. Looks like the furry models have mostly worse quality resources than the popular general models at Civitai. Or there is some other problem.
Stable Cascade sounds cool. It's awesome how incredibly fast this stuff develops.
I have to test more with the refiner switch, but usually it makes things worse. Looks like the furry
Can't say no to a cute otter like that! Thanks for making this tutorial. It finally encouraged me to to set up comfyui and automatic1111. Is there an easy way to get some info on generating 3D SBS images, or movies?
Can't say no to a cute otter like that! Thanks for making this tutorial. It finally encouraged me to