That doesn't really look like it respected the prompts too well, unless that was some checkpoint tha
That doesn't really look like it respected the prompts too well, unless that was some checkpoint that has such specific style and no loras.
AI tried regional prompter and while it worked somewhat with the horizontal/vertical divides it quickly failed when I tried it with masking, especially since I was trying to use loras for specific parts of the regions
Awhen I tried latent instead of attention mode (which is supposed to separate loras into regions) it rendered something that looked nothing like what I was getting withotu regional prompt or with regional prompt in attention mode
AAnd yeah, I'd rather avoid using inpaint since the goal is to generate lots of decently looking images straight out of the txt2img with maybe optional pass of batch img2img to refine it or alter the style somewhat.
UThe problem with my showcase is that my softedge was sloppy. I need to redo my open poses. I sorta patched things together in Photoshop, added characters together one at a time. My softedge is the problem in this instance for clarity.
Aso which extension is that?
UI have earlier versions that are much more crisp but they lack my jedi
ULatent couple.
UWith it I'm using a custom made mask as shown
Adoes it need any other extension to work?
UYes control net
Aaside from controlnet
UNope
Ayou sure?
UJust those two
UAnd Lora support o guess which is standard tho in webui
UThere might be a Lora extension needed to insert them into prompt with weights. I don't use auot1111 anymore
UI've been on a fork for awhile since I don't trust him to stay active
UI know he's been active AF lately. Competition good thing
Uheres an earlier step in the process
Ulater on, there's some things to figure out as i advanced
U
Uit does very well, these were made off open poses + softedge nets
Ui ended up ditching open pose when i made some spacing changes and added another character, now i need to go back and manually readd the open poses
AThat does look better
Unone of the detect open poses work very well on mulitple subnjects
AI was thinking of antoher extension that apprently need somethign else to work
AGitHub
This extension replaces the built-in LoRA forward procedure. - GitHub - opparco/stable-diffusion-webui-composable-lora: This extension replaces the built-in LoRA forward procedure.
Uyes that
Uthat i think is standard in the fork i use i think? idr
Ubut it lets you do prompts <lora:0.5>
Ustyle weighting
Aapparently it says it's compatible with Composable-Diffusion but I can't find what that is
Aand I swear I read somewhere that you need both for it to work
Ucomposable is already part of it all im pretty sure, its just standard, the bones of webui
Aif you don't use a1111 what do you use?
Awhatever is less buggy I'll take it 
Aspent last week trying to get it to work
Uits auto1111 derived so its more or less the same, def less buggy imo but no guarantees
UGitHub
Opinionated fork/implementation of Stable Diffusion - GitHub - vladmandic/automatic: Opinionated fork/implementation of Stable Diffusion
Uand it comes with themes baked in so you dont have to use the orange on black
UGitHub
Stable Diffusion web UI UX. Contribute to anapnoe/stable-diffusion-webui-ux development by creating an account on GitHub.
Uthis one is another good alternatiuve too
Uthe maintainers love their work and practice better coding best practices than auto
Uand compatible withall extensions
Aand now lycoris loras dont' work sometimes, just gives out an error: RuntimeError: The size of tensor a (9) must match the size of tensor b (4) at non-singleton dimension 1
Aany ideas?
AGot it to work for a bit by reinstalling but it shit itself again few generations later.
Usometimes it has to do with dimensions of images? always has to be a multiple of 8 for some things
A512x512
A512x768