Record & train action transformers

SuperVisual is a platform for crowdsourcing data collection and training action transformers.
Capture content & intent
  • Record audio visual content
  • Collect keypreses, clicks, selections
Language model integration
  • Prepare dataset for GPT-3, Flan
  • Replay & compare sessions
Crowdsourced data collection
  • Secure collection using tab sharing
  • Private local data storage