An AI-Generated Picture is Worth a Thousand Words...

Today, we’re excited to announce that we’re launching the beta version of one of our most highly requested features—visual generation!

As we previously promised we are adding several new features that will make the AI Dungeon Dragon experience more immersive, visual generation being one of these new additions. Players type what they want and the AI generates a unique image just for them.

The current model does especially well with locations. Here are some examples.

​Some types of images don't work quite as well, but this is just the first step towards higher quality visual generation. We're already working on the next iteration of visual generation which will be significantly better and also include more styles users can choose from.

Our plan is to eventually also add an auto-generate mode, where the AI will automatically decide what scene it should generate and create images for you, allowing for frictionless visual adventuring in infinite worlds.

This is an exciting step towards our vision of enabling anyone to make the next evolution of immersive experiences and we're excited to see what our community creates!

Questions

Who will get access to visual generation?

Visual generation will be a feature for Dragon AI users. Each generation will cost the same amount of energy as a Dragon generation.

How do I use visual generation?

To use visual generation you need to turn on "Enable Beta Features" and the New Game Screen in the settings page and then click on an action and select "Generate Image".

What model are you using?

The first model we're using is based on the awesome work of DALL·E Mini by Boris Dayma et al. We have several other models/methods that we're testing internally and hope to have available to users soon.

Can I generate NSFW content?
Right now the model won't be able to generate NSFW content. We are currently revising our content policy to include rules on visual generation.