🔅 A lot of time has passed since we launched the ability to generate images with artificial intelligence in
our RU chat and we have concluded - the generation is raw, the images are terrible and, in sum, our community is not happy.
🖼 What to do in such a case? Work on the issues! So meet...
👨🎨 The updated version of Stable Diffusion neural network code-named v2.2.2 XL Beta is already in our Bot.
📃 What's the difference? In quality. In the flexibility. In our pride in being able to properly integrate a neural network into our already tricky bot.
❓ So how do you create an image?
💬 Go to the
RU chat room and write
/нарисуй and your query text. The more words, the better the result. К
⌨️ An example of a valid query: young Disney socialite wearing a beige miniskirt, dark brown turtleneck sweater, small neckless, cute-fine-face, anime. illustration, realistic shaded perfect face, brown hair, grey eyes, fine details, realistic shaded lighting by ilya kuvshinov giuseppe dangelico pino and michael garmash and rob rey, iamag premiere, wlop matte print, 4k resolution, a masterpiece
👎 If you don't like the resulting version, you can now update your query by playing around with the settings at
/дорисуй 15,25,1, where the numbers are:
▫️ Scale (1 to 20) - how free the AI is to interpret the query and whether it should consider all parts of it. 2-6 is the minimum, 7-11 is half of your query, 12 and above is the whole query taken into account.
▫️ Steps (from 10 to 100) - how many steps the AI will take when generating. In theory, the more steps, the longer processing and the better result, but in practice this is not always true.
▫️ Method (from 1 to 8) - method of final image processing: k_lms, k_ddim, k_euler_a etc.
🗯 Visit our RU chat and try the new AI, but remember, the number of attempts per hour is limited.
