The original title is 10 words: "GPT-Live Needs an Interruption UI, Not Just a Microphone Button"

Short summary

OpenAI's GPT-Live voice models need interfaces that go beyond a microphone button to handle a full state machine including interruption, recovery, and permission states. Accessibility requires visual status indicators, restrained ARIA live regions for meaningful transitions, and an interruption test matrix covering scenarios like user speaking over the model, network changes, and background tabs.

• •Voice AI needs a full state machine, not just a mic button — covering interruption, recovery, and permission states
• •Accessibility requires visual status plus restrained ARIA announcements for meaningful transitions only
• •Interruption test matrix should cover user overlap, permission loss, network changes, and background tabs

Generated with AI, which can make mistakes.