How to make images with Artificial Intelligence with a few words, with ruDALL-E

Hi everyone! The “Prompt To Image” processes are blossoming everywhere on the web since Katherine Crowson presented the VQGAN+CLIP tool and made it public. This “Synthetic Imagery” (or GAN Art) was fantastic, but a bit difficult and slow to use.

You’ll find plenty of articles about this, and hundreds of “Google Colabs” with the code to play with. The result is often slow (about one hour to get an image).

The next step was to find similar tools (you enter a text, you get an image) on web pages like https://hypnogram.xyz, https://text2art.com/ or http://gaugan.org/gaugan2/ or the very easy https://www.wombo.art/ (have fun!).

There are tricks you quickly learn to use with each tool. Adding words to the prompt, like artists names or words like steampunk – here are bridges, a mantis, an owl, and for the first one “bird leather gold“:

Each site has its flaws, and one must use them to get things. For example, GauGAN2 is made for landscapes, so if you ask “Lake and forest” you get a realistic scenery. But if you ask “Totem” it’s lost, and there come the cool things:

The possibilities are infinite. Just give two words like “Airship Fire”:

Not what we expected, but good images, inspiring maybe if you write stories, poetry, or if you draw. Make 20 of them with automation and you’ll find a few great pictures.

oOOo

I made plenty of movies with these:

oOOOo

This year the Russians invented ruDALL-E ( https://rudalle.ru/en/demo ) and it’s different, more realistic, and MUCH FASTER than every other similar tools. It needs about 1-2 minutes to make one image.

The results are less “digital artist”, and much more realistic, because it’s trained on millions of photographies (an AI must be “trained”). This morning, today, I made a few dozens, like these 3:

Yessss possibilities are great. And you don’t have to write in Russian, they translate. Good.

oOOOo

This team made a BOT, which is on Telegram (yes, the app, it’s on your phone and your Mac, right?). You’ll find it on the page, it’s here: https://t.me/sber_rudalle_xl_bot

  • On this bot, you use the ruDALL-E Malevich (XL) Model, which is very powerful.
  • Each prompt gives you THREE images, you just have to save them on your computer, and it works on your phone too.
  • You have to prompt in Russian. Therefore you have to use a translation tool like Google Translate to invoke it.
  • If you find a good prompt, you can and must repeat it: each time you’ll get NEW images.

Here are images with the prompt “Airship in the mist”, which is “дирижабль в тумане”. I made 135 of the same prompt today. I’ll make a clip later. Here are 12 of them:

These are cool, right?

Here’s my YouTube channel with plenty of clips made with these: https://www.youtube.com/channel/UCkYi6dzJ5emaY0tPGat3k9Q

Have fun!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s