Installation

It's amazing. I haven't gone back to A1111 except to train TIs. Haha

It has somewhat decent mobile support, but you'll have to run it on a compatible PC and then connect to it over wifi or the internet. A easy way to access it online is run it on a PC and then install ngrok. Then in cmd prompt or terminal you run ngrok http 8188 and it will give you back a public link to use.

WAS-NS doesn't use ffmpy, it uses opencv-python-headless[ffmpeg] instead. So not sure why you're encountering that.

You have to run the requirements with the correct Python that ComfyUI is running on. So if you downloaded the portable version, you have to run the requirements with the python_embedded/python.exe executable.

Additionally you need cmake on your system and in your PATHS: https://cmake.org/download/

That doesn't help solve any issues. You need to provide the full stack error. Follow the installation instructions closely.

Something may be wrong with your ComfyUI install possibly. They're hard-coded options for the selection so should be select-able unless something is wrong with the front end.

@SeriouslyMike Yeah, WAS-NS requires things that comfyui uses, so can't install properly if ComfyUI is running.

@SeriouslyMike I thought I pushed to the node descriptions? I did have a PR or two between so maybe it got reverted by accident cause I had a conflict with my own changes.

It's basically an HR pass node. You can put in upscale factor you want, and then how many upscale cycles it takes. Which is how many times it's diffused at incremental steps to final factor.

You can hook up a secondary diffusion model and secondary cycle step that the secondary model is used at. So if you have 2 upscale_cycles you would set secondary cycle to 2. If you had 4 upscale cycles you'd set it to 3 to make it start halfway.

You can scale down the denoising for each step based on the cycle_denoise and enabling scale_noise. So for a fresh diffusion with no unit image you would have starting_denoise at 1.0 and then could use say cycle_denoise 0.5, and with scaling it would decause the denoise each step starting at 0.5 at second cycle step.

You can define use latent upscaling, if enabled it will use selected sampling mode tondo upscaling. This bypasses the VAE input and doesn't decode and encode the image between each pass. This will bypass upscale models and scale sampling FYi though.

You can select a scale sampling mode and use that alone for upscaling, which will supersampler results in raster for better upscaling without a upscale model.

You can hook up a upscale model to use the upscale model for incremental upscaling, scale sampling is used to downscale the result before the next cycle step.

You can hook up a processor model which is just a upscale model that is 1x output, meant for stupp like dejpeg artifacts, or denoising, etc. This will run between each step, and before optional upscale model.

You can add a additive positive and negative prompt and desire strength. It strength scaling is active it will exponentially increase the strength to the cutoff value. Helpful for transitioning to another prompt. It uses the Combine Conditioning (Average) node behind the scenes. This can be used for all sorts of stuff. For example you could have main model be anime model, and secondary be realistic model, then use the additive prompts to describe 3D modifiers and negative stuff like anime, art, oil painting, etc. This will have the effect of transforming the anime image into like a 3D image.

Sorry if any errors. Wrote this on my phone. Not at the computer.

Definitely join the Element community listed on ComfyUI GitHub. I share example workflows and answer support on WAS-NS.

Are you using windows? If so, I'd suggest downloading the portable package, which has the whole environment. Then follow directions for portable install. Clone it to custom_nodes, go into was-node-suite-comfyui folder, and then run `path/to/ComfUI/python_embeded/python.exe -s -m pip install -r requirements.txt` by changing the path/to/ComfyUI/python_embedded to the path to your python_embedded folder in your ComfyUI root.

https://github.com/WASasquatch/was-node-suite-comfyui#recommended-installation

@Heckerman I'd need to see the full error. Maybe cmake isn't available. You can make an issue on GitHub. There's an issues tab on link above.

@MysteryWrecked This is usually way with Python in general. ComfyUI Manager will get an update hopefully soon to handle installations better. But I am making a install.bat cause it's frustrating dealing with everyone else's frustrations on a daily. Lol

@MysteryWrecked I have pushed a install.bat that should hopefully solve the issues.

@Genie123 That and Right Click -> Open in Terminal

@MysteryWrecked Yeah, no I hear ya there. Things have changed dramatically. I struggle to keep up, and Web Development, my old field of work, has completely changed pretty much.

@EricRollei21 Node Manger can't really install nodes that do anything involving packages that are running. Numpy, etc. You need to follow my instructions to install. Namely running requirements.txt through your embedded_python/python.exe.

ConfyUI Manager is only going to work for nodes that don't really need dependencies or use totally different dependencies from what ComfyUI has already imported. It's a cool tool but should be a stand alone tool to actually achieve what it advertises.

@EricRollei21 you need to run through python_embedded/python.exe in root of ConfyUI if you are using portable version

Install the nodes according to the instructions. You need to run the requirements with ComfyUI's python environment. So for portable you need to run requirements through python_embedded/python.exe

No

There's supposedly a way to capture the widget info and use it elsewhere, but I think there is an issue with custom widgets not being supported.

Yeah, that's cause it's very old and before the Utils -> Primitive node. You can just use that and input into a seed input and it'll transform into a seed.

Thank you! I appreciate the kind words!