Stable Diffusion WebUI, actually unstable...

[astronaut954]

I know this discussion may not get much attention from the community, but this is important to talk about. Please don't see this as an insult, I'm only trying to point out things that would make the tool better:

1. Why Stable Diffusion:

Stable Diffusion is something that tries to make a better alternative for people compared to limited tools like Midjourney. Its main purpose is to give users more autonomy over AI generative images. People daily contribute to this tool and invest a lot of time so it can compete with big companies with overpriced plans and no ethical concerns about AI use. This is only possible if people USE the tool and yet Stable Diffusion WebUI fails at the very basic user-friendly features it needs.

2. Stable Diffusion WebUI Unstableness:

2.1. While I'm not an expert with Python, Git, and so on, I'm also not some completely lost newbie, yet I have come across a LOT of problems while trying to use Stable Diffusion WebUI. I constantly got my files denied access, extensions would crash my entire Stable Diffusion WebUI, /temp/ folder wouldn't clean itself, extensions would constantly break the program, reinstalled countless times the application, and had to constantly search for troubleshooting, had to constantly manually install package managers, and so on. One proof of all those problems is the constant issues that are being reported in the current repository.

2.2. Since I'm not a developer (yet I have a good background about it) I can only speculate about the problems. Since the program is not compiled it supposes that the user has the exact scenario as the one that is uploading the repository, so if there is even a slight problem with Python, variables, deprecated extensions, updates of extensions, Python, and package managers it will make everything suddenly stop working. This is a pretty bad scenario for every type of user, advanced, intermediate, or beginner, it will take a lot of people's time and sanity to fix it.

3. Stable Diffusion WebUI (un)friendliness:

3.1. One of the first things that got my attention, not in a good way, when using Stable Diffusion WebUI was its, dramatic pause, UI... The UI has some issues with basic gestalt and UI principles that could be easily fixed. I'm certain that Stable Diffusion loses something like 50% of its users just with this first bad impression, an organized and better-looking UI would help so much the community withholding much more users. What makes me even more confused is that on the official Extensions page, you can get a better Theme, that doesn't make much sense, it's just one more step that will make a lot of users quit.

3.2. Another important aspect is when we talk about UX, that's certainly a much more complex topic and I can't provide a good answer to this problem now. For sure Stable Diffusion is not only an alternative for paid AI generative image, but also a tool for more advanced users that won't get what they need with other tools available there, yet for sure it is possible to combine simple UI and advanced UI fulfilling both types of users needs. For example, include a button for "Easy Mode" and "Advanced Mode". Again, I know this is more complex and maybe there isn't any consensus about this.

3.3. The last point is that while I know this is a GitHub repository using Python this is probably the most problematic point about Stable Diffusion WebUI. Since Python needs to get downloaded and most users have little or no idea how to do it they just will quit and when they don't quit they will get across another core problem: Stable Diffusion WebUI incompatibility with the latest Python version. That's probably one of the worst scenarios that I can imagine, even for advanced users that's a total pain in the a** since you need to have more than one version and maybe even other variables in your system to that old version.

3.4. Why there is no shortcut inside the Start Menu or even an .exe file to run the program? It would make it much more friendly and act as a proper program.

Sorry if for some reason at any given moment, my text seemed rude, my main purpose is to help with the right questions and answers.

[BetaDoggo]

Stable-diffusion-webui isn't really meant for casual end users. There are plenty of alternatives that fill the "commercial image-gen but local" niche, fooocus for example. Most of the stability issues you mention are the result of the high extensibility of the webui. Extensions can and do break things but many of them couldn't work without this risk. The python issues are inherent to python, there's not too much that can be done about them.

I think the ui itself is as basic as it can get without limiting user control. Most of the settings are set to sane defaults which will generate an acceptable image, users can play with them if they want to.

[astronaut954]

Yes, I understand that the tool isn't designed for casual users, and the things I pointed out reinforce this statement. Yet a better UI and more stable version would benefit everyone, even the developers since it would be less susceptible to problems that affect every workflow, and advanced users wouldn't need to constantly troubleshoot the program.

For example, the fact that the program doesn't have a compiled version makes it even more unstable, since it is in Python it is a better alternative for developers and at the same time a compiled version would benefit all types of users. I am almost sure that the UI can be more user-friendly and at the same doesn't limit the user, for example, you have a simple mode and an advanced mode, that would benefit everyone.

I guess the main problem would be: it would take more time to do it and updates would take longer. That's for sure some drawback, yet I think the benefits are huge.

[light-and-ray]

1. Stable Diffusion Webui is a webui for stable diffusion, obviously. If it restrict user ij settings, it isn't complete ui for stable diffusion, it's something else which uses sd in kernel

2. Python version is not a problem for advanced users. Just write python3.11 instead of python in command line

3. You can make binaries using pyinstaller. But why it's better then bash/cmd script?

Automatic1111's is very automatic software for this kind of software. You just do git clone, and it is all. "Stable" in stable diffusion name is not connected with webui. Stable diffusion is a project by CompVis: https://github.com/CompVis/stable-diffusion Is doesn't have any UI