Generative AI · 2021 to 2024

AI Video and 3D Experiments

Testing new tools by making things with them

When a new generative tool comes out, the best way I know to understand it is to make something with it. These experiments cover AI painting, style transfer, AI video and text to 3D, and range from a few minutes of playing around to a short film made for a 48-hour challenge.

October 2023. A short film made entirely with AI for Runway's Gen:48, a 48-hour film challenge. Stills from Midjourney became video in Runway's Gen-2; the sound design was sketched in Stability AI's Stable Audio; the voiceover script came from ChatGPT and the voice from ElevenLabs. The theme came from picking up go-stop again, a game played with hanafuda (flower cards), and from having less than 24 hours to get going.

Be Good: a holiday horror trailer

It started as a joke, a "what if?" story a friend posed. I made it on my phone in under an hour with Luma AI's Dream Machine, Midjourney and CapCut's audio libraries, as part of Luma AI's Creator Program.

Text to 3D with Luma AI's Genie

A livestream generating 3D models from text with Genie, then cleaning them up in Blender to prototype new scenes in minutes.

Turning my friends into Arcane characters

Using open-source machine learning models to make stylized images and videos in the look of Fortiche's Arcane, and what that could mean for artists.

Can AI paint like Bob Ross?

Racing Bob Ross with NVIDIA Canvas. It took a little longer than 27 seconds.