Skip to content

Comment on AI and the Future of Pixel Artparent

Comments

Dreambooth extension for automatic1111 just came out. Can be run on CPU even. Haven't tried that extension myself yet, but I have followed a youtube video last week and I trained it using a colab in a few minutes (of training - getting the hang of the whole process was maybe 20 or 30 mins including watching the vid?). I think dreambooth is already perfectly within reach of regular users, and is already being used for... ehem... questionable purposes that took 1000's of images and days and days of tweaking and training for 'traditional deepfakes' just 6 months ago. The pace of advancement is breathtaking, it's literally impossible to keep up even if you spend 100% of your time on this.

Thanks for the tip about automatic1111 with dreambooth!

Do you happen to have a link to that YouTube video you followed?

It was actually two now that I look in my bookmarks: https://www.youtube.com/watch?v=w6PTviOCYQY and https://www.youtube.com/watch?v=FaLTztGGueQ . That first one is about 1 month old and was already about running Dreambooth locally. So to be perfectly honest I'm not really sure any more which info I got from which video. Probably most from the second one. It's worth it for just the thumbnail - although I haven't been able to get results at that level yet, holy smokes is it impressive.

Not OP, but this is fastest tutorial for getting dreambooth up and running in automatic1111 I’ve found:

https://youtu.be/_GmGnMO8aGs

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.