More visual interactivity with the cats #1
Description
right now there is a single image for each cat. and that has a small role in interactivity, since it is shipped up along each query to help add "flavor" to the response from Gemini that might be based on the cat's personality.
but we could also adapt this image, perhaps every few API calls, or in response to changes in sentiment from the user. we could send the image, along with a prompt such as "modify this image to make the cat look spooked", to the Imagen API.
considerations:
- doing it less frequently may be necessary, since this is a more expensive query
- maybe we cache or store the results and recirculate previous results
- could we have users append labels to rate the resulting images and create a pareto front of modified cat images for each sentiment/cat combination?