Home
Artists
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register
Partychan Telegram Find&Fuck Waifus ThePornDude
Home Artists Posts Import Register

New preview on SD 0.5 (Patreon)

Published:
2022-09-13 18:47:21
Imported:
2022-09

Content

Tried to put a bunch of quality of life improvements that was suggested here.
Keep in mind that most the options will improve with time, for now i'm trying to find the quickest way to implement improvements, so it don't slow down development that much.

I just need to pretty much finish implementing gfpgan and this update should be ready.

Changing model will not be simple in this update, but it should be possible. (Download the model on the web and drop in a GUI folder)

Output model now save the path, tooltips if you hover the mouse on a box along other stuff.

I will try to finish it today or tomorrow, but there is always the chance of a script being problematic and taking more time to fix it.

Files

Comments

Anonymous

Great work man, looking forward to it :)

Anonymous

Don't know if anyone else has reported this, but I have found an annoying bug. If you create multiple samples with random seeds, it records them all as having the same seed in the filename and the text file. This means that promising results can never be refined or reproduced, which is a real issue for me. Thanks.

Anonymous

Found a better optimization if you can integrate the changes into GRisk GUI. I don't know if it's feasible but you could have an option that determines how many chunks the model is broken into. Maybe 1024x1024 could be possible on a 4GB GPU? "generate 512x512 images from a prompt using under 2.4GB GPU VRAM in under 24 seconds per image on an RTX 2060" Link: https://github.com/basujindal/stable-diffusion

Anonymous
>>91651657

I've noticed that too. I think dev knows seed code "was a mess" they said. I'm guessing it will improve.

Anonymous

This looks great - can't wait! I'm not sure where to submit suggestions, but maybe down the road, you would consider adding Variations? Afinetuned model is being worked on by Justin Pinkney:

Anonymous

Sorry, in case you wanted a link: https://twitter.com/Buntworthy/status/1561703483316781057

Anonymous

Im loving this tool so far, its so useful to generate ideas quickly and so easy to setup as a executable and not dependent on cloud, what a dream. There are 2 small quality of life functions I would love ,if you could consider the suggestions . 1.) Hotkey/shortcut to render and cancel :: A Hotkey like "ctrl + enter" to start render and "ctrl + C" to cancel render, so that we can quickly iterate on prompts from the text box without moving to mouse click to start a new render. 2.) Before and After preview (img2img) and prev render preview :: a thumbnail or side by side preview to see your original comparison image and your previous render*(with text under it to show seed and prompts) , side by side with the new render , to help see before and after.

Anonymous

It would be great if you could also add the ability to animate the generated images with a "thin-plate spline motion model". https://github.com/yoyo-nb/Thin-Plate-Spline-Motion-Model

Anonymous

Would like to see another button on the bottom alongside Render, Cancel All Renders and Skip Current Render. An appropriate label would be: Save & Continue Functionality: Stop rendering the current image, finalize it, save it and move on to the next render. It would complement Skip Current Render very well as users could alternate between the two instead of having to complete all steps when they see something they want.

Anonymous

Would it be possible to have multiple cards running in parallel to speed up the processing?

Anonymous

I wish it would run with my GTX 1660 at home. It's a pity I can only do stuff on my PC at work.

DAINAPP
>>91698548

It should work with a 1660, what problem are you having? Send me a PM

DAINAPP
>>91686036

Once the GPU selection is ready, just open the .exe twice and select a card for each interface.

DAINAPP
>>91683873

Let me see if I undestand, is the skip button, but it save the result so far as a png. Is that it?

DAINAPP
>>91678169

This probably would need a entire new interface, it could be done in the future, but it will take some time.

DAINAPP
>>91672358

1) Good idea, it is now on my todo list. 2) This is a good idea, but I think it can be better, once 0.5 is released gonna play around with some ideas for this.

DAINAPP
>>91667682

Been wanting to add this since it was released, but still did not had the time to play with it, but definitely want to add it to my gui.

DAINAPP
>>91656874

Wont chunks break the image generation? It would generate 4 heads in a 1024X1024 image. Will take a look on this github, thanks!

DAINAPP
>>91651657

This is actually a limitation on how the seed currently work, will fix it very soon

Anonymous
>>91667682

That is so awesome! Really appreciate everything you're doing!

Anonymous
>>91683873

Correct. It's to save time when you want the result without having to finish the full render. It would be useful for me anyway.