I tried to create a new study group page with undecided content under the name “aaa study group” and found this page
- I changed it to a dated page because it was probably made and left unused.
2022-11-07 - Stable Diffusion Study Group (2022-09-30) continued
-
[NovelAIDiffusion] - Using round brackets in NovelAI is pointless.
-
HyperNetwork
-
11/3 “NovelAI Aspect Ratio Bucketing” released under MIT license
-
11/2 Runway Frame Interpolation
I can use pictograms at the prompt.”
-
I don’t think there’s an abundance of training data, so I’m just falling back on Unicode character descriptions.
-
1995 Brian Eno: https://www.wired.com/1995/05/eno-2/
-
In the future, you won’t buy artists’ works; you’ll buy software that makes original pieces of “their” works, or that recreates their way of looking at things.
- In the future, instead of buying the artist’s work itself, we will buy the software that creates that artist’s new work. Or to put it another way, software that reproduces the artist’s way of seeing things. (Translated by Nishio)
-
Fine-tuning a Diffusers-based StableDiffusion model (U-Net+Text Encoder)|Kohya S.|note Stable Diffusion is additionally learned with “Irastaya” - TadaoYamaoka’s Development Diary.
Promptia Magazine
“by artist”
Runway「Infinite Image」
- outpainting
- https://runwayml.com/
@umiyuki_ai: tool to make image cropping and tagging (DeepDanbooru) useful. Could be useful for hypernetwork learning and such. https://t.co/HCqhAn5yBS
- Danbooru did it in a public forum and was the subject of bashing, so the trend is for individuals to do it locally.
This page is auto-translated from /nishio/2022-11-07勉強会 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.