Help wanted

[sebaxakerhtc]

Successfully adapted original code and now with pleasure using it through auto1111's webui, but I'm not a python dev. Need help with:

• [x] using checkpoints from default sdwebui path
• [x] using VAEs from default sdwebui path
• [x] using LoRAs from default sdwebui path
• [x] randomizing seed
• [x] SD 1.5 Support**
• [ ] from previous: SD1.5 adds objects at standard settings. Need fix.
• [ ] imageslider (gives error, so using gallery) conflits with sd modules
• [x] autosaving outputs to sdwebui outputs/demofusion folder
• [ ] interrupting process
• [ ] live preview and progress bar
• [x] allow to set custom resolutions** (for now the largest width/height is automatically sets 1024, 2048, 3072
• [x] Solving problem with image padding using img2img
• [ ] from previous: fix SDXL non-square images. SD1.5 works fine
• [ ] Escaping 77 tokens

[Jaylen-Lee]

That's why we're considering re implementing based on k-diffusion instead of using the original code. Including multifusion and mixturediffusion, their original versions are based on diffuser implementation similar to demofusion and need to be rewritten for webui. 😉

[sebaxakerhtc]

That's why we're considering re implementing based on k-diffusion instead of using the original code. Including multifusion and mixturediffusion, their original versions are based on diffuser implementation similar to demofusion and need to be rewritten for webui. 😉

Yeah, I know. But, hey! I already added support for choosing checkpoints from default WebUI path in testing branch! Will add vae support soon. Also works on SD1.5 adaptation. But just try how it works! Originally ❤️

[w-e-w]

• autosaving with infotext see how save_image_with_geninfo is used https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/b372fb6165c2b0f3990d9a3ff06245d25358c0b8/modules/images.py#L530 for ouptput path see search for opts.outdir_txt2img_samples you can also create you own settings page

• interrupting make a button of some kind that runs interrupt() https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/b372fb6165c2b0f3990d9a3ff06245d25358c0b8/modules/shared_state.py#L79 check for shared.state.interrupted is true in you pipline, if true then stops or you could Implement somehing that is completely separate but serves the same function

  the reasion to use the one in webui it means that buttons on other tabs will will also work for you tab

• live preview this is more complex you will need to make a api endpoint that will returns the preview image then when when it's generating with JavaScript the browser call you preview api and update the image viewer

• allow to set custom resolutions this is already implemented by webUI, see ui-config.json search for "DemoFusion/Height/maximum": 8192,