WebUI: Ideas for improvements

[softtaco1]

WebUI implementation ideas to improve flow

There have been many new features/models in the works and releasing lately. Eventually, it could result in a cluttered WebUI. Here are some proposals to better future proof the UI:

1. Drop-Down Model Select: Instead of toggle-able models at the top of the page, having a drop down menu to choose the base model. (Flux, SDXL, SD 1.5) Upon changing models, refresh Lora/Embedding lists to show only those compatible with specified model.
2. Save “Presets” which saves exact parameters: Gives the user the option to save exact parameters in all tabs to help with their particular workflow. These parameters include things such as CFG, Resolution, ControlNets and ControlNet image references. Once a preset is saved, it is saved under the parent model and checkpoint selected, meaning if a user created a preset for Flux it would not show up under SD 1.5. The same would happen as well if they saved a style for PonyDiffision— it would not show up under JuggernautXL. Presets could be exported/imported, assisting in following guides and workflows. They can be copied to other models/checkpoints if the user decides to as well.
3. Simplify Resolution: Based on model selected, allow a drop-down menu to choose premade popular aspect ratios designed to work with that particular model. On the backend, these have premade resolutions for that model. For example, selecting 1:1 aspect ratio for a SD 1.5 model would generate an image in 512x512, while selecting 1:1 aspect ratio for a SDXL model would generate an image in 1024x1024. In the drop-down, there is an option for “specify exact resolution” which then allows you to choose something different if so desired. For img2img, this would work the same way. HiResFix would default to “resize by 1.x” This would help as surely it’ll start to become hard to remember which models should be at which resolutions.
4. NEW feature- Send to new model: After image generation, there is a feature to where you can send that exact image that was just created to a new model via IMG2IMG/Inpainting. It works similar to using the info tab and dragging a photo in and then specifying where to send the photo, but is simply a bit faster and easier to use. As some models have strengths where others are weak and vice versa, this would be helpful to tweak an image made in one model with a different model. Once it’s sent to IMG2IMG/Inpaint, the user could specify a preset as well based on the model that they had saved to better optimize their workflow.

These are some simple ideas to optimize the flow of Forge! Feel free to implement absolutely any of these ideas if they sound useful at all.