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

Hello all (Patreon)

Published:
2022-12-07 01:54:48
Imported:
2023-01

Downloads

Content

I was rushing to complete the next build of SD today, but ran out of time. Almost sure will be ready tomorrow.


Before writing more about it, just want to share that the new model for Rife-App is almost ready (check video)

This model definitly is the best one yet, and is still not finished, so it should be a little better than the example video above.
Here the original footage before being interpolated by Rife-App.
https://streamable.com/nvk8yw

Extra test: One Punch Man:

Original 

Interpolated

Now more about SD:
There is a bunch of small modifications on this next version.
A lot of those modifications will be invisible for you all.
The thing is, the more complex a code get, the worse is to work on it, slowing becoming a spagetti code. As this happen, bugs get easier and easier to appear, build progress get slower and it becomes hell to keep updating the code.
So, I took some time to clean and improve the code, to help me keep working on it on the future. Invisible for all, but it help me.

Now for some real modifications.

  • Selecting V2 model will only show schedulers that work on it.
  • .PT files will convert .chpk files correctly. 
  • .PT files will work in any future version of the GUI. Currently the .PT could be broken with an update to the GUI
  • You can select the folder with all the models. (Getting ready to add models as separated downloads)
  • You can save the images as .png or .jpg
  • Show some new alerts, like selecting img2img without a origin.
  • On training, can select on how much steps display a render. Or disable it.
  • Select resolution for training.
  • Small bugs fixes.

Comments

Anonymous

any news on the new version yet thanks.

cool1

I have a question on dreambooth training (related to the next release too since I don't know if that will be different in the part I'm asking about). In the GUI 0.58 (I don't have the "hack to run 2.0" version yet since I couldn't download it), it allows you to enter the "Exclusive Prompt for Images". In the models\dreambooth\(nr of training steps) folder after training is an args.json file that has things like: "class_prompt": "style", - that's shown twice - once in the concepts list bit and once outside it It also shows "class_data_dir": "./dream_train/style" (though I don't see that folder - maybe because I downloaded the lite version of 0.58) but the GUI doesn't seem to have anywhere where it asks for the class prompt. A youtube video showing another dreambooth training UI shows a field where they enter an instance prompt and a class prompt and it also has a field where you can enter an optional classification data folder (for prior preservation). How is this GUI determining that the class prompt is "style" without it being input in a prompt in the GUI? Is it currently hard coded to set it to that? Would it be better if it had a field in the UI for it and maybe also an optional field for the folder for the class data? eg. in case the class isn't a style eg. it could be a person or an object or something.