My team is developing a game surrounding AI-generated images: Chat TPG – The Prompt Guesser. One player will generate an image using AI tools, and the other players will try and guess what prompt was inputted into the tool to produce that image. Our game prototypes might answer the following questions:
Should people come up with their own prompts entirely, or should it be pre-selected by the game?
- We want players to be able to exercise their creativity, but also come up with prompts that are some what standardized in difficulty. A prompt that is more concrete like “dog on the moon” would likely generate an image much easier to identify that an abstract prompt like “happiness”. Thus, it is important that we find the right balance between creativity and structure, so that the game is enjoyable for both the prompt generators and guessers.
- We have played around with different ideas of how to test this. One prototype may allow players to come up with prompts entirely on their own. Another may allow players to pick prompts from a deck of cards and then plug it into the tool. A successful prototype will likely be somewhere in the middle: perhaps there are a set of 3 categories (e.g. adjective, person/thing, place) and then the prompt generator can either come up with their own word for each of these categories, or pick from each of three decks aligning with each category (e.g. adjective –> “happy”, person/thing –> “dog”, place –> “moon”). There is also a world in which the prompt generator picks a slightly less specific card that allows them to exhibit more creativity, while still giving them boundaries to work within (e.g. adjective –> “emotion”, person/thing –> “animal”, place –> “planet”).
- I think that having the player select a category card (e.g. “emotion” “animal” “planet”) within each pre-selected category (e.g. adjective, person/thing, place) will standardize the difficulty across guesses, and create a nice set of boundaries for both the prompt generator (to come up with a prompt quickly) and the guessers (they know that the clue only contains 3 main words, and what categories they are in). I feel that these boundaries will strike a balance between difficulty, standardization, creativity, and fun.
When will the game end?
- Currently, there is no set outcome of the game. Likely whoever guesses closest to the actual prompt gets a point, but when does it end? It is important to set these terms so players know what they are working towards, or how much time they have left to defeat their opponents.
- One prototype could have the game end after a certain number of rounds (e.g., each player gets 3 turns where they are the prompt generator), and another could have the game end after a player or team receives a certain amount of points (e.g. 10).
- My guess is that rounds would work the best, because in the other option, the player who has most recently been the prompt generator will be at a disadvantage, as they did not have an opportunity to get any points during that round.
Should this game be played in teams, or individually?
- Currently, there is no clear incentive for the prompt generator to provide either an easy or a difficult prompt. Thus, the prompt generator does not have a clear motive or goal, since in the round where they are the prompt guesser, there is a not a clear way for them to earn points for themselves.
- One prototype could test the individual model, where players win points for themselves every time their guess is closest to the actual prompt. For the team model, perhaps the prompt generator could generate a prompt for the opposing team. This incentivizes them to make the prompt more difficult. Teams could get a point for each word in the prompt that they guess correctly (up to 3 points if they guess the adjective, person/object, and place correct).
- I am guessing that the team model will work out better, because there is a clear competitive aspect to each round: the prompt generator vs the opposing team.
