Andrey Chudnovskiy

Mneme

In progress

A native app for low-latency, full-duplex voice interaction with OpenAI’s Realtime API, built to run from a shared client architecture across iOS, watchOS, macOS, and visionOS. The project explores the engineering challenges behind making voice AI feel immediate, fluid, and native on Apple platforms.

Screenshots

Technical Highlights

  • Integration with OpenAI Realtime API for low-latency voice interaction.
  • Generated API client from OpenAI’s OpenAPI spec using Swift OpenAPI Generator.
  • Real-time voice streaming across iOS, watchOS, and macOS.
  • Shared cross-platform client architecture spanning iOS, watchOS, macOS, and visionOS.
  • Built around full-duplex interaction rather than turn-based request/response UI.
  • Exploration of native interaction patterns for voice-first AI software.

Coming Next

  • SceneKit-based AI visualization.
  • Further refinement of cross-platform interaction design.