Storyboarding with AI

I'm heading into developing a clickable prototype for SYL (Shine Your Light), the AI-powered mindfulness adventure game I am co-founding with Sylvana Levy.

I confess that I have been leaning on my tried and true brainstorm partner, ChatGPT, to help me flesh out the user journey, and so unsurprisingly, I began with Open AI's DALL-E 3 to begin storyboarding the prototype.

Image generators have come so far, so fast that it's a bit dizzying, and the DALL-E 3 integration into ChatGPT is no exception. In fact, it was so good, that I wondered how the other image generators I had tried in the past might have evolved.

The starting point of our adventure is in a mysterious forest. A path branches in two and the player must choose a direction. Here is the prompt I gave to all the image generators I could think of (with the exception of Midjourney, because, well, I already have too many AI subscriptions...):

A dark, misty forest path where the trail forks into two shadowy directions. The forest is dense with tall, gnarled trees that have thick, twisted trunks and bare, sprawling branches reaching up to the sky. The atmosphere is eerie and mysterious, filled with fog and soft blue-gray lighting. The ground is lined with moss and ferns. The central point of the image is the "V" of the fork in the path, leading deeper into the forest. The trees loom over the paths, and their canopies reach the top of the image, creating a tunnel-like effect. No characters, animals, or artificial elements—just nature. Mood is somber and immersive, like the beginning of a moody fantasy adventure.

Style: Semi-realistic, cinematic lighting, concept art quality

Aspect Ratio: 16:9

Color Palette: Cool tones – deep blues, charcoal greys, muted greens

Detailing: High detail, atmospheric depth, realistic tree texture, soft fog diffusion

Here are a few of the results

With the same prompt, only two gave me branching paths: DALL-E 3 and Gemini. Meta AI, Stable Diffusion, OpenArt, Flux.1 and Da Vinci (among the others I tried), stubbornly refused to branch no matter how I tweaked the prompt. The why remains a mystery!

After more experimentation, I have ended up going with DALL-E 3, which so far seems to be able to generate image after image in the same world. That said, they all come out very dark, and need some tweaking with Photoshop.

Finally, and just for the fun of it. I used the same prompt to generate a 360˚ environment with Skybox AI

All I have to say, is that this experimentation is such a blast!

#genai #dalle3 #davinciai #leonardoai #flex #metaai #chatgpt #ai #userjourney