SwarmUI

SwarmUI copied to clipboard

Allow loading same model twice in memory for refiner, avoid speed penalty when using <base> and <refiner>

2

### Feature Idea When using lora in a prompt, it's necessary to use `` and `` to limit the lora effect to only the base: ``` woman, solo, sitting, chair,...

jtreminio

### Feature Idea In the Model Downloader tab - When selecting Base Model under Model Type, the model gets saved to the stable-diffusion model folder. I currently see no option...

saintbrodie

### Feature Idea To conserve space and help with organization, it would be nice to have the option of when Starring an image it was moved to the Starred folder...

CaptainVarghoss

### Expected Behavior Wanted to set default VAE for user from server admin panel without need to log-in as this user from other tab with non-localhost IP. ### Actual Behavior...

Derfuu

### Feature Idea **The Problem** Currently, the image generation process in SwarmUI involves a generation pass (txt2img) followed by an upscaling pass (the second pass). While this is effective for...

Mooshieblob1

### Expected Behavior When I add any `` tags to a prompt to perform regional inpainting, and enable the refiner panel, I expected the Generator to run the refiner, as...

nonplayer

### Feature Idea Hello! Hopefully this isn't a too-bothersome request. There are probably a couple of ways this could be improved, but basically, I'm requesting an improvement to the search...

nonplayer

### Feature Idea One of the features of SwarmUI is to retain the last used weight of a LoRA was, and if the LoRA is chosen from the LoRA tab,...

brendanhoar

LoRA Weight value from grid parameter are applied to one and only one inline <lora:...> entry when the currently processing LoRA grid parameter is a blank/empty value in the list

1

### Expected Behavior The Grid's Lora Weight parameter values should never be applied to inline loras found in the prompt, even if the grid's Lora parameter values include an empty/blank...

brendanhoar

### Feature Idea Some users are running gens using horizontal and vertical values that are simply swapped. E.g. the outputs are meant to simulate a physical camera or smartphone's output...

brendanhoar