As we kick start 2024 we wanted to look back fondly at some of the events which shaped our world in 2023. We hope you enjoy our first issue of AI Pixels below.
Our 'AI Pixels' comic I've shared with you here was created with Dalle.
Before the festive break we were doing a lot with image generation. We explored different prompting and meta-prompting techniques to achieve compositions that were able to meet expectations, especially when the image needs to convey something specific. We'll tell you more about our experience as the project progresses but, one thing that I've found can work really well, is the 4W1H prompt structure set out by Christine C.
Essentially, you're breaking down your prompt into different elements which describe 'what, who, where, when and how'. 'AI Pixels' used the following prompt structure. - 'What' described the category of image (a comic book in this case) and general composition - 'Who' set out the characters that should appear in the image and their relationships to each other - 'Where' described the setting the characters should be in - 'When' described aspects of the image/ interaction such as time of day - 'How' gave general art direction. In this case, using a seed parameter we use for our image generation at Tomoro and some notes on how the image should be rendered. It's a useful framing to help steer creative input and reduce the likelihood of having to re-generate images that don't meet a (relatively) specific need.
If you have any questions feel free to discuss further with us here.
