python optimizedSD\optimized_img2img.py --init-img input/test.jpg --prompt "cat,fusion" --H 640 --W 512 --ddim_steps 70 --n_samples 4 --n_iter 3 --strength 0.3

2 10

Might fuck around and use this as an init for something

0 3

I then used this image as an "init", an image from which another image is synthesized, resulting in new images with similar compositions, but different details.

Can't remember how many times I did this, but it was a lot. Probably in the ~50x range 5/

0 9

Disco Portrait Diffusion 1.0 - Release

Hi, here is a DD notebook I was configuring optimized for my model Portrait Generator.

Portrait Generator V2 + Init Images included.

Colab: https://t.co/EtT2kFKUVf

8 27

"The search for the final midnight orb to save humanity hidden in shallow waters they prowl and scavenge the murky waters." 🌊

using an init image (last image) that was created with the same prompt using an earlier AI algorithm. 👾

0 0

You all remember the I've been giving updates on? Well, I was running a section of it through stable to get a defined style and I left the init strength on a little low, and it gave me a surprising output I immediately was like, YES, to. 2/?

0 7

introducing more to the prompt and reducing the strength of the init image. You can see the result still bears resemblance to the original, but the image synthesis starts to take over more

0 9

img2img(下絵画像から画像を生成)

python scripts/img2img.py --prompt "ここにワード" --init-img "ここに下絵画像のパス" --strength 0.8

strength は自由度で、低いほど下絵に忠実になり、高いほどランダム性が高くなる(1.0まで)

失敗する場合は、画像縦横サイズを256,512,1024などにするとよい

0 7

Good evening all:)
with an init image.

2 6

Reworking an early piece as an init to (Original, plus new versions)

0 11

I can now mess with init images, though I think I have no idea what I'm doing yet...

0 2

Quack it’s the init! Stop ngeatbquacking community, come join the and see for yourself my quacker, it’s ducking awesome quack

13 18

画像はレンブラントの「夜警」をサイバーパンクにしたかったやつ。
Init Image Settings:のskip_steps:を多い数にするとlist index out of rangeエラーになってしまう。
10とか30でないとできなかった。画像はskip_steps:30。
解像度上げたいのに〜〜ぐぬぬわからない

0 0

Image to image explosion of the library after feeding the generated image as init.

9 85

Used the init image on the left to generate the picture on the right with prompt: "Anime girls sitting on the hills, beautiful background, makoto shinkai, thomas kinkade, james gilleard". Turns out image to image works really well!

3 33

Disco Diffusion(no init)> Photoshop >Disco Diffusion >Photoshop >Topez Gigapixel

2 11

The Photoshop liquify tool, dodge/burn tool and copying pieces of imagery from other parts of the piece to modify (such as the golden eyes from flowers) were all employed to “rebuild” the face, which bears almost no resemblance to the photo in the init image.
4/5

1 5

Gm gm ⭕️ my 🍿 using as my init for creating magic in

1 6