LiveCue Bloglivecue.co
← All posts

2026-06-20 · LiveCue Team

On-device speech on macOS and Windows

How Apple Speech and Windows AI Speech keep interviewer audio off the cloud — with Whisper as a 99-language fallback.

LiveCue layers speech recognition so you get low latency without shipping audio to the cloud on supported hardware.

macOS

On macOS 26+, we use Apple SpeechAnalyzer via a native Swift bridge. Audio never leaves your machine for transcription.

Windows

On Windows 11 24H2+ with Copilot+ hardware (or capable CPUs), we use Microsoft.Windows.AI.Speech — the same on-device stack documented by Microsoft.

When we use the cloud

If your interview is in Hindi, Arabic, or another language native engines handle poorly, LiveCue falls back to Groq Whisper through our managed API — metered, not stored.

This is the layered strategy from our pre-launch architecture: native first, cloud when it actually helps.