DAINAPP
patreon
Would you use a GUI for Textual Inversion?
🕑 Added 2022-09-09 20:10:22 +0000 UTCTextual Inversion is like that, you take 5~20 images of something and train a variable that can be used on Stable Diffusion. This image explain it well.
It seen that it require 12Vram to train it (did not test myself) and a few hours of training.
There still a LOT of stuff that I want to do before even thinking of doing a gui for this, but would like to know if anyone would like to use it? Keep in mind the vram required for it.
This is a tool that eventually I will make, but it may get a bigger priority depending on the answers.
Comments
Javier García
With Google Colab, you can pay for better GPU there
I mean, there's probably more porn images than any else out there. My guess is, whenever see a version of this, it's going to produce some amazing results, lol
coldas icefree
ya lol
GRisk
There is probably quite a few people doing that lol
Somebody train the model on porn and share it ;-)
It would be great to do some custom training for items that are missing from the data set. The default set has no idea what a killer whale is!
Thank you for the reply. Any other way to accomplish such task, with some kind of online tools, resources? Any ideas or suggestions? Thank you.
GRisk
Train or Fine Tune SD on CPU would probably take years. Unless you have thousands of CPU to use.
Would love to have this as well
I would like to use this functionality, but the VRAM requirements goes beyond my GTX1070 (8GB VRAM) specifications. Is there any other methods we can use to train Stable Diffusion, perhaps via CPU only? Or with some available online tools?
Would love to see this included.
The original model intentionally excluded some visual references. I'd like to add some back in.
This would be fantastic as many said. Would be great to add in new characters that aren’t already in the system. Such as new video game characters or anime characters that come out.
this would be awesome for defining very specific artstyles, instead of adding a bunch of artists names to the prompt and praying :D
That would be awesome!
