sseramemes stable diffusion

sibelius / ccsseraphini

cc @sseraphini

https://ccsseraphini.vercel.app/

174 stars 50 forks source link

sseramemes stable diffusion #1236

Open sibelius opened 1 year ago

sibelius commented 1 year ago

[ ] be able to generate a meme using stable diffusion + text interpolation
[ ] https://huggingface.co/stabilityai/stable-diffusion-2

nephi-dev commented 1 year ago

gonna leave some refs here:

Stability just dropped SD v2

SD2.0 -> https://github.com/Stability-AI/stablediffusion

The colab to Dreambooth(fine tune memes in it)

fast DB -> https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast-DreamBooth.ipynb

The colab to Textual Inversion(fine tune subjects, styles, anything. Less effective as DB)

TI -> https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/sd_textual_inversion_training.ipynb

sibelius commented 1 year ago

we can use v1

nephi-dev commented 1 year ago

it's easier(and has more implementations than the v2), I also sugest using it, 1.4 or 1.5, both are good

sibelius commented 1 year ago

how would you combine image + text ?

like this https://imgflip.com/memegenerator

nephi-dev commented 1 year ago

maybe using PIL to interpolate both, it has a func to add text to an image(on a specific pixel), maybe the hard part will be detecting where it needs to go, or just selecting the pixel by yourself should be enough