Thank you! John found UnityEyes2 (aleksmcfly) and trained off it alongside some data the main developer for this curated, but we want to use actual captures!

Our goal is to use 40 people to create around 1200 minutes of training footage we can use to make things more robust on some serious hardware! We are going to use around 5 headset models to do this alongside a varied enough demographic to account for ethnicities often ignored with these solutions (Chinese, Japanese, anyone with monolids really alongside pale and dark skin)