Realtime AI Fun

One thing that baffles me about a lot of gen ai tools coming out is how low they set their ambitions. This tech is exciting and capable of much more than creating videos that, let's be honest, look better when they're made by actual people. Where's the innovation? Where's the unique user experience? Where's the FUN?!

With a foot firmly in the fun camp, here's an experiment with realtime AI. A simple webcam and AI gestural recognition are pumped through a combination of stable diffusion and ChatGPT to create and endless mix of realtime, interactive animations. Despite never touching a screen or a mouse, the user has control over the content, the colour and the movement all with simple hand gestures.

Media Pipe turns my webcam image into a silhouette and reads my body movement and hand gestures. Making the V sign with my hand randomises the prompt from a list of pre-selected keywords. ChatGPT reads the prompt and generates a colour palette to match. And Stream Diffusion creates images guided by the silhouette, the palette and the prompt.

This tech is still quite new, but I'm seeing it as potential component in next generation experiences in entertainment, retail, culture and earning spaces. And trust me, it's really fun!

Just below this you'll see another experiment combining AI tools.  Here, the time and weather are queried in six cities. The time controls the hands of the clock. And the time and weather information is parsed into a text prompt and fed into Stable Diffusion, which then returns a design appropriate for the city and the weather.