The Hands-Free Desktop Kit — ChatGPT Voice on Desktop: Setup, the 5 Phrases, the Cost Math & the Privacy Checklist
The companion to the full video on ChatGPT Voice landing in the desktop app (July 23, 2026 — macOS + Windows, Plus/Pro/Business/Edu/Enterprise). Voice doesn't just answer anymore: powered by GPT-Live it can start, check, and steer real agent work in ChatGPT Work and Codex while you keep talking. INSIDE: (1) WHAT SHIPPED + HOW IT REALLY WORKS — the two-loop design we confirmed by inspecting the packaged app itself: a live conversation loop, durable worker threads, and the explicit handoff (with transcript context and a stable ID) connecting them — plus why the silence is designed: only done/blocked/needs-you gets spoken. (2) THE 2-MINUTE SETUP from OpenAI's own docs — surface picker (Work vs Codex), the Voice control, and the exact macOS permissions for screen context (AppShots). (3) THE 5 PHRASES WORD-FOR-WORD — start, check, steer, screen, parallel — each mapped to an officially listed ability, plus the 'Wait until I ask you to respond' trick. (4) THE TWO-METER COST MATH — connected Voice time (~6 credits/min on business plans) is billed separately from the tasks it starts, plus the full plain-Voice limits table for every plan and the budget rule that keeps the meter honest. (5) THE PRIVACY CHECKLIST — 30-day audio retention, delete-vs-archive, the training opt-in rules, and the honest gotchas (no video/screen-share in plain Live at launch, one conversation at a time, custom GPTs still use the old voice). Every claim verified against help.openai.com on launch week. Independent field kit from Hyperautomation Labs — not affiliated with OpenAI.
Free. No spam. Unsubscribe anytime.