Voice input for Codex

Speak the task; text lands at your terminal cursor. Approvals too, while you read the diff.

System-level input — text lands wherever the cursor is · zero integration

No plugin needed: task descriptions, follow-ups and approvals in Codex CLI are all typed — now they can all be spoken. Transcription runs on-device; neither voice nor code leaves the machine.

At a glance

DimensionSpoketypeNotes
IntegrationNo plugin — works out of the boxText lands wherever the cursor is
Works inTerminals, editors, any input fieldNo per-app setup
Transcription runsOn this computerAudio never leaves the device
HotkeyCustomizableDefault ⌥Space (Mac) / Ctrl+Alt+Space (Windows)
PriceFreeNo account, no purchases

Codex is an OpenAI product; this page describes using Spoketype alongside it and is not an official integration.

In an approval loop, voice beats keys

The Codex loop is “describe the task → read the diff → give feedback” — three text-input steps, and three places your wrists pay. Spoketype makes all three speakable: describe it, review it, respond to it.

Three postures that just work

One — dictate task descriptions with full context in a breath. Two — speak review comments while your eyes stay on the diff. Three — with parallel tasks, speak into whichever terminal you switch to.

When it pays off most

Anyone running agent tasks all day whose approval comments have numbed their fingers.

FAQ

Does it need a Codex plugin?
No. Spoketype is system-wide input — text lands wherever the focus is.
Is my code or voice uploaded?
No. Transcription runs on-device; audio never leaves it.
Does it work in Codex sandbox mode?
Yes. Spoketype only types text at the cursor and doesn't interfere with the sandbox.
Is it free?
Free — no account, no in-app purchases.
What about the browser-based Codex?
Supported — browser input fields are just another place the cursor lands.
Download Spoketype Free

macOS & Windows · No plugin · Audio never leaves