Thib∞dさんのイラストまとめ

Some early tests on a new workflow.

With longer prompt:

"arcane style, portrait of Emma Watson"
Euler_a / cfg 7 for A - 9 for B-D

A) Dreambooth on SD 1.5 - 10K steps - lr 5e-7 - constant
B) DB on SD 2-768 - 3K steps - 1e-6 - constant
C) DB on SD 2-768 - 3K steps - 5e-7 - constant
D) DB on SD 2-768 - 3K steps - 5e-7 - polynomial

I'm training dreambooth model with SD-768.
Results seems to overcooked too fast.

75 refs images
1000 regs
3000 steps

Prompt: "portrait of santa claus, arcane style"

Santa Claus.
#stablediffusion finetuned V2-768

@magifactory @KaliYuga_ai I tried 512 and 728 training.
With 512, I have not tan solid output but results are very bad. I tried 3K steps and 11K and result are awful.

Example with v2 - 512 at 11K training steps (left) (with diffusers)
old training with v1.5 (right) (with shivam repo)

4/6 : Guillaume IVERNEL (Chasseurs de dragons)

Lot of startups disrupting some markets with big seed funding.

Markets:
- 🎮Video Games (with startup like @Scenario_gg)
- 🎵Music
- 🎲Entertainment globally (I'm on it, stay tuned)

💲More than $1bn in Seed / Series A

@StabilityAI 🦄

2/2

Cat on mushrooms?
#stablediffusion