DreamStudio Mastery: Beautiful AI Art with StableDiffusion

StableDiffusion is the software behind DreamStudio. There are other ways to use it as well.

The first steps into DreamStudio: select the right model for your output and describe what you want to generate in the prompt.

How many pictures, format of the pictures, configurations of the algorithm, Seed.

The model is the basis of the picture generation, consisting of millions of saved pictures that determine the output. So, different model equals different output.

You like what you got and want to have more of it? Here is a good possibility.

A quick walkthrough through your account settings.

Without a good prompt, your pictures will be lacking purpose. In this class we dive into the depths of Prompt Engineering.

Now we follow down that road and develop a system to create better prompts.

With a negative prompt you exclude all the things you don't want to have in your picture like two heads or 6 fingers.

Here is my negative prompt for copying:

Blurry, oversaturated, low-res, grainy, black and white, out of focus, poorly drawn, artefacts, signature, watermarks, title, disfigured, ugly, deformed, bad anatomy, mutation, mutated, disgusting, mutilated, mangled, surreal, gross proportions, bad anatomy, extra limbs, missing limb, floating limbs, disconnected limbs, malformed limbs, poorly drawn hands, malformed hands, two heads, extra fingers, mutated hands, too many fingers, fingers, 7 fingers, 2 fingers, 3 fingers, 4 fingers, 6 fingers, blurry hands, blurry fingers, long neck, long body, two bodies, two heads, zero heads, poorly drawn face, cloned face.

Inpainting lets you select part of your picture and the AI would then generate a new version of that part.

I loved the face, but the top of her head was strange, so I impainted it.

What are the problems with inpainting and when should you generate new pictures instead?

Your picture is too small? Missing something to the right or left? Outpainting helps. Unfortunately, we don't have a solution in our Stable Diffusion eco system, so we move to Dall-E for this task.

Save your prompts in your browser, share them with others, learn from other prompts!

We use an external tool from Tencent to improve the quality of the faces. It's not magic, but it transforms a mediocre picture into a good one.

It doesn't work for whole pictures, but you can animate the faces of your Stable Diffusion creations, quite a thing, in my opinion - unfortunately this tool from MyHeritage is not totally for free.

Stable Diffusion doesn't know how to handle text and one sometimes gets quite strange results. Therefore we add text with another tool, may it be PhotoShop - or much easier, with Canva.

You can generate simple headshots with Stable Diffusion, but there are tools out there that are easier to use.

If you don't want to develop your prompts yourself, then this is a great tool - take a look at it, because at the least, it can give you some inspirations. And Promptomania looks really great, that for sure!

Where there is something public, there is a lawyer. In this class i talk a little about the law situation with Stable Diffusion.

I wish you the best for your AI art projects, hopefully with a lot of help from DreamStudio! Thank you for taking this class ?

As Jackie Chan does, we all do - and love it: Outtakes, FAILS!