Inferred Image Descriptions
The easiest way to get started with Image Generation is to simply let the system use the text of each response to decide what to illustrate this turn. This doesn't require you to do any specific setup in the Definition, it will simply look at the Character response and generate something based on that.
Depending on your context, this may lead to Characters generating inconsistent images or images that don't align to you particular Character or story. It's hard to predict what it might use in the returned response. We encourage creators to use the Image Style attribute to establish some consistent themes in this case.
In Sakura Grove Curse, the standard text response is considered when generating an image, rather than establishing a Direct Image Description. To allow it to maintain some consistency across images given all this content, a few context setting terms are added to the Image Style:
Last modified 4mo ago