How to use ChatGPT to create a refined prompt

Are you familiar with the expression "being afraid of the blank page / white canvas"? I must admit that more than once I felt like that when I tried to think of a prompt to use with one of the numerous AI Art Generators I am creating images with. Well I discovered a method for myself which helped me to deal with that kind of feeling and I want to share it with you. 

1. PromptSilo
PromptSilo is a webpage where you can find all kinds of information to put into a prompt. It gives you ideas for a prompt. For example you want to create an image which depicts a landscape with certain plants or flowers. It gives you a list of flowers, trees, whatever you like. You want to create a painting of this landscape in a certain style - just click on the artist.

This is what I got:
 bouganvillea trees style of Edgar Degas daylight

2. ChatGPT
The second step is to ask ChatGPT to refine the prompt. Here is the refined prompt:
Capture the graceful essence of bougainvillea trees in the style of Edgar Degas, bathed in the soft glow of daylight

3. Submit the prompt to an AI Art Generator (in this case: Bing Image Creator)

Here is the image created with the above prompt.

Once you have a refined prompt you can change it easily. You can replace the kind of tree you want to see, you can change the style, you can change the light. The possibilities are endless!

Changing the seed numbers in Stablecog AI

 Stablecog AI is one of the AI Art Generators I use quite often. A few days ago I watched a video about the importance of Seed numbers (the video referred to Playground AI, but the information given is valid for all AI Art Generators.). I do not want to go into detail here, I just want to give you some examples for the impact of varying the seed number.

The prompt for this image was:

dwarf holding a tulip in his hand, digital painting, cinematic photography

Seed number: 4

There is a dwarf, and he is holding a tulip in his hand. However to me the stem of the tulip does not look right, and the dwarf looks like thinking: "What is this? It looks dangerous to me!"

Instead of changing the text of the prompt I changed the Seed number from 4 to 44.

Well, the dwarf is holding a tulip in his hand and the tulip definitely looks better. The dwarf himself looks more relaxed. But there is this long tulip on the right side of the image which seems to grow out of the dwarf's back. Of course one could always think of removing that unwanted item digitally, but it would cost me a lot of time.

So I changed the seed number again, this time from 44 to 55.

And this time I was quite pleased with the result: The dwarf looks relaxed, the tulip in his hands is an eyecatcher because of the orange yellow colours, and the tulips in the foreground form some kind of a frame.

So if you are not pleased with the first results of your prompt try to change the seed number! In most cases you will get something usable out of this process!

How to organize your prompts with Google Keep

I also got a category like "sports" in Google Keep!

For quite some time I have been looking for an efficient way to organize the prompts for my AI images. I've got loads of them - some from ChatGPT, some from other AI artists, some of them made up myself. I like to try out the same prompt with different AI Art Generators, and I like to play with prompts, just change a bit there and there. That means I wanted to find a way to get access to them very quickly.

Using notion to store my prompts
In the first months of my AI Artist career this was a quick and efficient way. However the page got longer and longer, and the longer it was the longer it took for me to find the prompt I wanted to find.

Using Google Keep
In Keep you can tag your notes. The first tags I used for my prompts were "ai" and "prompt". But this was not efficient enough. So I thought of the way I organize my AI images on my hard drive: There I use categories like "landscape", "people", "surrealism", "objects" only to name a few. I decided to use the same categories for labelling my prompts in Google Keep, with one additional tag: "style". I apply this category for prompts which describe a certain style - like expressionism, or a certain artist and such. 

Creating Mockups for Your Zazzle shop

For quiet some time Zazzle is providing some Mockups together with the product you have created. However Zazzle encourages to create your own mockups, be it that you take photos with the product itself (which requires that you bought the product), or creating mockups using online mockup generators. 

Smartmockups.com
Smartmockups.com is a website which offers a lot of mockups - for mugs, t-shirts, cards and a lot of other products. Say you want to create a mockup with a mug. Select "mugs" and then "free" and you get all the free mug mockups you can use.

Which Zazzle products do you want to promote via Mockup?
Creating mockups is not very difficult, but it is time consuming. So I go through my store and look for products which I think could do with a bit of promotion on the one hand. Then I go to smartmockups.com and look which products offer the most free mockups.
After that I go to the product, go to "download" and download the design. This is the file you have to upload to smartmockups.com in order to place it on a mockup.

Edit your designs on the Mockup!
You can edit the design on the mockup - enlarge it or move it. Try to make it look as similar to the real product on Zazzle as possible.

Upload the mockup image to Zazzle
I have a free account on smartmockups.com which means that the images I download from there are relatively small. It could be that Zazzle requires a bigger image. However, there are several free photo enhancers on the web. Use one of them and then try it again. 

All in all, the process of creating mockups is time consuming, especially if you are just beginning to do this. So, as I said before, focus on the products you want to promote!

And here is one of my Mockups. You will find the mug here.

Nightcafestudio: Working with DALL-E Part 2

 Okay, you are on the Nightcafe website and you want to create an image with DALL-E2. Click on "Create". This will open a page with all the possible creation methods: Stable Diffusion, Dall-E2, Coherent, Artistic and Style Transfer. Today we will generate an image with DALL-E2, so click on the image with the cool green chairs. 

Now you have to insert a text prompt. Try to describe the image you want to create as precise as possible.

I chose: "Pixar style 3D render of a penguin wearing a tophat, 4k, high resolution, trending in artstation"

Now you have to decide how many images you want to create. Each image uses up 1 credit, so you may want to save credits by going for 1 image. But: the one image you go for may be awful. So I recommend going for 4 images, and then selecting the one you like best.

I got those four images:

All four penguins are cute, but I decided to publish the one in the lower right corner.

DALL-E2 does not have additional parameters to influence the outcome of the creation process, like Stable Diffusion has. There is only the text prompt. This can make it more simple, but also more complicated.