#6 - Subject detection and instruction
While it is possible that the model will detect what you're looking for if you describe it plainly, precise details will help provide a boundary and goal for the model to achieve.
Once you've established that goal, you can then elaborate on what you want to happen. It is possible to direct the model to place subjects in an entirely new scene, and further to provide guidance on what they should be doing in that scene. This method can be used to create a variety of images where a character consistently is involved - something that is currently very hard for standard text to image models.
In this example, the prompt includes:
- instructions for detecting a certain number of characters
- create a new scene where
- the characters are all doing the same thing (sitting), which is different from the start image
- each of the characters is specifically prompted to do something unique.