The Process
For my second story my goal was to create interaction between characters and place them in a coherent scene using a single prompt, without having to put the scene together in Canva manually.
I was able to create a first draft of the video in Leonardo.ai by uploading an image I created using ChatGPT and using it as a reference image to create a consistent character. I then entered prompts for the scenes and interactions I needed for my story. I was able to complete the process but the end result felt wooden and inflexible. I was unable to vary the expression of the character, her placement in the scene, her poses and her interactions with other characters. The result was monotonous, over-the-top "cute" and ultimately not credible.
I then discovered how to build my own custom GPT, provide it with instructions and a knowledge base of reference characters and an art style. This was a breakthrough as I was then able to quickly generate all the images I wanted for the video in a way that achieved my vision for the story.