Nick Fisher

Hi all - I wanted to showcase the current state of AI text-to-motion models, so I put together this small demo app [0]. It uses the MoMask[1] model on the backend (I'm not affiliated with that team, just a fan of their work). They have a visualization on their HF space[2] but I thought it would be better to see the generated animations with an actual model. I'd really like to take this further...

Last week I built a little experiment to live-stream ChatGPT-generated, foreign-language dialogue to Twitch. Here's the result! It's not flawless, but for this task it works very well. Admittedly there's slight survivorship bias, as I do skim the results and discard anything that stands out as immediately bad. That's actually pretty rare though, and as soon as the ChatGPT API is available,...

Rotobot is, hands-down, the quickest way I've seen for semantic segmentation. I'm not a pro visual effects designer, but I was blown away by how simple it was to get going. Pros: Pre-built semantic classes! Makes it very easy to use. Cons: Sparse user documentation