pydn
Advantages of this solution
What are the advantages of using this vs running the workflows via ComfyUI API? Are there any time savings / speedups / memory savings?
I think this is mainly if you want to run image generation in for example a non-gui setting (think server env). Also, this is freaking awesome. Do you know how tedious it is to create complicated comfyUI workflows in pure python? Literally weeks of time saved using this.
What are the advantages of using this vs running the workflows via ComfyUI API? Are there any time savings / speedups / memory savings?
You can quickly experiment with your configuration in ComfyUI and export it to a Python script for integration into your development pipeline for production use. This typically involves generating over 3000 pieces per day, for example. Additionally, you have the option to include extra features such as prompts, numbers, etc., to customize the output based on your preferences or at the client's request.
The main concern for me is that this repo seems to be stale now. No commits since Aug 2023.
Thanks for your input everyone! @maximilianwicen and @aetherwu did a great job articulating the purpose of this extension.
@trevordunn This solution is pretty stable so doesn't require a ton of upkeep unless there is a major change in the structure of the main ComfyUI repo. I'm keeping an eye out for major changes and pushed a patch today (with the help of a contributor) to keep up with some breaking changes in the latest changes to the ComfyUI repo.
Hi @pydn, this project is awesome but I'm not sure I'm using it correctly. The README says "Now you can execute the newly created .py file to generate images without launching a server." but when I go to my ComfyUI folder and execute "python myworkflowscript.py", ComfyUI still boots the server (and that takes time).
Not sure I'm doing everything correctly or if this is how it is supposed to work. Cheers and thanks for this project.
Hi @niko2020, that is how it's supposed to work. It means you don't have to launch the server in a separate terminal before running the code. In the future, please raise a new issue if your question does not relate to the issue you are commenting on.