Loading

The Great S-Cape the difference a little txt makes

As an aside, the work of Jonas Peterson, whose website link is provided below is worth looking at to see what an accomplished CGI artist and photographer can achieve in this media area. WARNING: do not try this at home 😏

This is the third and final 'shot in the dark' covering the loose ends from the previous posts. Text prompts and 'style' choices play an important part in the output presented to you. At most txt to image sites there are extensive FAQ's that can help to orient you and thus your students to enhance the quality of you experience when using AI. Some; like Mid Journey, that run on Discord servers take some getting used to whereas Stable Diffusion Web app, Dall-E and Dream by Wombo are intuitive enough to facilitate student use. I understand that Adobe have entered the foray with Firefly. Firefly is now available as a public beta and can be accessed with your normal Adobe login.

The following links are accessible through DoE servers

I don't work with txt prompts as they are conventionally used and because I upload my own work as source images the txt prompts function somewhat differently in terms of how they influence the generation of images.

If your students or yourself are experimenting with Wombo you will be able to upload an image into the web app via the link above. The image will be decoded. The neural network will use the decoded image data combined with the decoded txt data to generate new data based on machine learning which will then be encoded back into an image format.

In the free Wombo web app you only get one iteration of any txt prompt / image upload. However despite the limited formatting options, it's enough to get the idea of how things work.
Stable Diffusion Online is a txt to image web app. If as a teacher you wish to explore possibilities that are more interesting you can go to the Dream Studio Beta, create an account and have a pretty interesting time.
MidJourney will not be available to you if you are using DoE servers.

Same txt prompt, different source images.

Same text prompt. No source image.

Txt prompt for the above images.

A highly detailed airbrush painting using this parametric equation to create captivating art, x(t) = A * cos(a * t) * cos(t) + A * sin(b * t) * cos(t) and y(t) = A * cos(a * t) * sin(t) + A * sin(b * t) * sin(t), where 'A', 'a', and 'b' are constants, and 't' ranges from 0 to 7π.

If the AI of your choice has options for negative prompts use this to correct problematic elements of the render. For example; hands, four fingers and thumb only. Feet, five toes only.

Most will have a bank of commonly used negative prompts. If not search Google. ‘Common negative text prompts’