Here are a sample of the outputs from three different AI models, Midjourney, Disco Diffusion and Dall-e to show how they can be used to create aesthetics and assets on a theme.
The models were given prompts including different combinations of the words dancers, cables, mobile phones, child, woodland and dystopia. The selection below aren’t searches or ripped off someone else’s work, they are original images, compositions and styles created through the AI’s interpretation of the language prompts.
DALL-E 2 – offers more literal and illustrative imagery, sometime with really good compositions.
Midjourney – abstract mood board or background imagery
I used Midjourney’s images to create an animated short by editing together different variations it created on the themes above. Asking it to create variations of a base images means it keeps the approximate composition – the prompts all included ‘in the style of anime’ to keep the images reasonably consistent:
Disco Diffusion – an open source tool using the same underlying AI algorithms as Midjourney. It is more of a general tool, and the Midjourney and DALL-E images were more coherent, but it offers a lot more options for creating animated worlds or scrolling landscapes, including using source images, so the fiction could start as a real place.
It can also create fast, disconcerting zoom transitions between images with different concepts, which would make the whole room feel like it is moving between scenes!
Or it can be used to just animate a scene – this could be slowed down to be almost imperceptible by the audience, very gently moving from one emotion/scene to another.
